Introduction

Drosophila melanogaster larvae have become a favored model organism to study principles of sensory perception (reviewed in (Gerber and Stocker 2007; Melcher et al. 2007; Apostolopoulou et al. 2015; Joseph and Carlson 2015; Rimal and Lee 2018; Widmann et al. 2018; Thum and Gerber 2019). It has been shown that larvae actively use sensory cues of different modalities to navigate through their environment. Thermo-, photo-, chemo-, and mechanosensory information is perceived and processed by the larvae’s peripheral nervous system to optimize their behavior output for favorable conditions (e.g. to find food sources) and to avoid unpleasant or even harmful situations (e.g. predators and deterrent or toxic substances) (Tracey et al. 2003; Gomez-Marin and Louis 2012; Apostolopoulou et al. 2014; Klein et al. 2015; Ohyama et al. 2015; Scholz et al. 2015; Apostolopoulou et al. 2016; Choi et al. 2016; Croset et al. 2016; Ni et al. 2016; van Giesen et al. 2016; Humberg et al. 2018; Tastekin et al. 2015).

An increasing number of studies investigating the neuronal and molecular basis of larval sensory perception has even enabled the identification of the respective sensory cells and receptors for distinct stimuli. Among others this includes a distinct set of olfactory receptors (odors) (Fishilevich et al. 2005; Kreher et al. 2005), as well as other receptor genes, like Gr43a (fructose) (Mishra et al. 2013), Gr28a (ribonucleosides and RNA) (Mishra et al. 2018), Gr33a, Gr66a and Gr97a (quinine) (Apostolopoulou et al. 2014), Gr33a, Gr66a and Gr93a (caffeine) (Apostolopoulou et al. 2016), Ir76b (amino acids) (Croset et al. 2016), Ir25a (denatonium) (van Giesen et al. 2016), ppk11 and ppk19 (low salt) (Liu et al. 2003; Alves et al. 2014), ppk23 and ppk29 (pheromones) (Mast et al. 2014), painless and ppk (noxious heat) (Tracey et al. 2003), ppk and dpiezo (noxious mechanical stimuli) (Zhong et al. 2010; Kim et al. 2012)), dCIRL and nompC (mechanosensation) (Scholz et al. 2015; Scholz et al. 2017; Yan et al. 2013), Ir68a, Ir93a, Ir25a, and Ir21a (temperature) (Klein et al. 2015; Ni et al. 2016; Hernandez-Nunez et al. 2021), and Rh5 and Rh6 (vision) (Humberg et al. 2018). This sensory description is complemented by a large-scale electron microscopic reconstruction of the larval brain (Winding et al. 2022). The reconstruction characterizes at the cellular and synaptic level the sensory inputs and initial information processing steps for the olfactory antennal lobe (Berck et al. 2016), the optic neuropil (Larderet et al. 2017), the gustatory subesophageal zone (Miroschnikow et al. 2018) and the nociceptive and mechanosensory brain centers (Ohyama et al. 2015; Jovanic et al. 2016; Takagi et al. 2017; Burgos et al. 2018; Jovanic et al. 2019; Masson et al. 2020).

Despite these advances, the structural basis of sensory perception in Drosophila larvae is far from fully elucidated because electron microscopic reconstruction is still limited to the central nervous system. A comparable analysis for the entire peripheral nervous system is not available.

Still, most knowledge on the ultrastructure of larval sensory organs and their sensilla derives from studies of different dipteran species carried out mostly during the 1970s and 1980s (Chu-Wang and Axtell 1971, 1972a, 1972b; Thurm 1964; Campos-Ortega and Hartenstein 1985; Dambly-Chaudière and Ghysen 1986; Denell and Frederick 1983; Hertweck 1931; Honda and Ishikawa 1987; Lanfranchi and Belcari 1990; Whittle et al. 1986; Zacharuk 1972; Keil 1997). Most of this work was done on different developmental stages of dipteran larvae using different methodical approaches. Current attempts to understand the relationship between structure and function of larval sensory organs of Drosophila are therefore hampered by an anatomical description that, on the one hand, is detailed for specific aspects and species but, on the other hand, has gaps for the Drosophila larva that are only indirectly addressed from anatomical results of other fly species.

To overcome this limitation and gain precise knowledge on peripheral sensory organ ultrastructure, we have recently analyzed the anatomy of the terminal organ (TO) of the Drosophila larva, its major external taste organ (Rist and Thum 2017). This was possible by taking advantage of technical improvements in volume electron microscopy. In particular, we used focused ion beam scanning electron microcopy (FIB-SEM) to gain precise, three-dimensional reconstructions of each of the 14 external sensilla of the TO. FIB-SEM can automatically generate serial images of ultrastructure with superior z-resolution compared to other common volume EM techniques. Extremely thin layers of the specimen are ablated by an ion beam and an image is taken by the SEM after each removed layer (Helmstaedter et al. 2011; Peddie and Collinson 2014). A more “classic” approach is serial section scanning transmission electron microscopy (ssTEM). In this approach, ultrathin sections of the complete sample to be examined are created, which are then scanned using transmission electron microscopy. These methods have increasingly been used to obtain 3D representations of cellular and even subcellular structures at high resolution (Peddie and Collinson 2014; Titze and Genoud 2016).

Using partial larval 3D volumes based on FIB-SEM and a full larval body volume established via the ssTEM technique (serial sections imaged with a TEM in scanning mode) (Schoofs et al. 2023), we have now analyzed the anatomy of the remaining three major head sensory organs, the dorsal organ (DO), the ventral organ (VO), and the labial organ (LO) at ultra-resolution. The three peripheral sensory organs are known to be formed by several sensilla (Hertweck 1931; Chu-Wang and Axtell 1971, 1972b; Kankel 1980; Singh and Singh 1984; Python and Stocker 2002). The DO was proven to be the primary larval olfactory organ based on anatomical, molecular and functional experiments (Chu-Wang and Axtell 1971; Singh and Singh 1984; Heimbeck et al. 1999; Oppliger et al. 2000; Python and Stocker 2002; Fishilevich et al. 2005; Kreher et al. 2005). Its prominent “dome” houses 21 olfactory receptor neurons organized in seven triplets that respond to different sets of odors. Less is known about the six peripheral sensory sensilla and their additional roles in thermosensation (Klein et al. 2015; Ni et al. 2016; Hernandez-Nunez et al. 2021) and putatively mechano-, and taste sensation (Chu-Wang and Axtell 1971; Singh and Singh 1984; Python and Stocker 2002). The VO and the LO are comparatively small sensory organs and have been little noticed in larval anatomical or functional studies (Chu-Wang and Axtell 1972b; Python and Stocker 2002; Miroschnikow et al. 2018). They are assumed to serve a mechanosensory and/or gustatory function.

Another focus of our work was to describe the morphology and ultrastructure of the external sensory organs of the thoracic and abdominal segments. A comparatively simple organization was reported for these. The majority of them consisted of only one sensillum, which was described to be either of the campaniform, basiconic or trichoid type (Kankel 1980; Singh and Singh 1984; Campos-Ortega and Hartenstein 1985; Hartenstein 1988; Campos-Ortega and Hartenstein 1997). Other studies, however, called campaniform sensilla papilla or pit sensilla. Trichoid sensilla were also called hair sensilla. For the basiconic sensilla, the nomenclature is most diverse, as these are called koelbchen, knob, knob-in-pit, hair-type B, sensory papillae, dorsal or ventral pit, or black sensory organ (Hertweck 1931; Dambly-Chaudière and Ghysen 1986; Singh and Singh 1984; Rist and Thum 2017; Singh 1997; Kankel 1980; Lohs-Schardin et al. 1979; Hartenstein 1988; Sato and Denell 1985; Lewis 1978; Campos-Ortega and Hartenstein 1985). In addition, not all sensilla are described and named in the fused first head and last abdominal segments (Schmidt-Ott et al. 1994; Courtney et al. 2000; Wipfler et al. 2013). Accurate classification and nomenclature of the different types of sensilla throughout the larval body – as applied in this work - will therefore be useful for future anatomical and functional studies.

Results

Proper classification of sensilla requires investigation of their external and internal morphology. However, the ultrastructure of scattered and small insect sensilla like that of Drosophila larvae is challenging to be investigated. Fortunately, recent advances in EM technique made it possible to image large regions of tissue, like the entire central nervous system of larval (Ohyama et al. 2015; Schlegel et al. 2016; Carreira-Rosario et al. 2018; Miroschnikow et al. 2018; Winding et al. 2022) and adult Drosophila melanogaster (Zheng et al. 2018; Scheffer et al. 2020; Schlegel et al. 2023), Caenorhabditis elegans (White et al. 1986; Cook et al. 2019), Ciona intestinalis (Ryan et al. 2016) and the larva of Platynereis dumerilii (Verasztó et al. 2020), as well as parts of brain of Pristionchus pacificus (Bumbarger et al. 2013; Hong et al. 2019) and rodents (Denk and Horstmann 2004; Helmstaedter et al. 2013; Motta et al. 2019). In this data we reconstructed the external sensory sensilla in a full body first instar EM volume (Schoofs et al. 2023) and in volumes of single sensilla of third instar larvae obtained by FIB-SEM. For single sensilla imaging, accuracy and precision are required which are often difficult to realize technically. In order to achieve these technical demands, we took advantage of FIB-SEM in combination with an optimized preparation protocol. This allowed us to exactly target even the smallest sensilla on the larval body wall for subsequent serial slicing and imaging by FIB-SEM (for details see supplementary Figure S1).

Structural organization of Drosophila larvae

The body of Drosophila larvae is divided into segments (Figure 1A): a pseudocephalon (Pce, from now on called “head”, Figure 1A’), three thoracic (T1-T3) and nine abdominal (A1-A9) segments (Campos-Ortega and Hartenstein 1997). The head is the strongly reduced head capsule of the larva, dorsally fused with and partially retracted in the prothorax (Courtney et al. 2000; Wipfler et al. 2013). The last abdominal segment, A8, is more appropriately named “anal division” as it is formed by the fusion of at least two abdominal segments A8 and A9 (Figure 1A’’).

Figure 1:

The largest sense organs of Drosophila larvae are located on their head (Figure 1A, inset left bottom). The DO and the TO are prominently located on the tip of the head lobes (Chu-Wang and Axtell 1971, 1972a; Singh and Singh 1984; Heimbeck et al. 1999; Oppliger et al. 2000; Python and Stocker 2002; Fishilevich et al. 2005; Kreher et al. 2005; Rist and Thum 2017). The VO and the LO are smaller and their ultrastructure has rarely been studied (Hertweck 1931; Kankel 1980). The VO is situated ventral to the DO/TO complex on the front of the larval head hidden behind rows of cuticle cirri (Figure 1A’, Figure 8B). The LO is inconspicuously located below the larval mouth opening (Figure 1A’, Figure 9B). In this work, we describe the morphology of the DO, VO, and LO each in a specific result section and present a conclusive nomenclature of their sensilla (Figure 2-9, Table 1). The detailed morphology of the TO was described in our previous work (Rist and Thum 2017). The anatomical data is therefore not shown again. However, its morphology is part of Table 1 to complement the description of the external head sensory organs.

Figure 2:

Table 1:

Sensory organs located on the thoracic and abdominal segments are of simpler organization mostly consisting of only one sensillum (Figure 1B). Different names have been given to these sensilla (Table 2). In the present work, we identified three basic types of thoracic and abdominal sensilla and refer to them as papilla (p), hair (h) and knob (k) sensilla (Figure 1B, 10, 11, 13, Table 2). This nomenclature corresponds to the one of Dambly-Chaudière and Ghysen (1986), which is most consistent with our findings. Please note that we have translated their term "kölbchen sensilla" into English as "knob sensilla". The term knob or knob-in-pit sensillum was already used in previous studies (Singh and Singh 1984; Rist and Thum 2017; Singh 1997). It allows us to use the same abbreviation (e.g. vk for ventral knob sensillum), which also allows for comparison with these studies. We are aware that the term papilla sensilla is questionable, as the outer and inner morphology rather resembles the one of the spot sensilla that we found in the TO in our previous work (Rist and Thum 2017). The term papilla sensilla is based on findings in light microscope data and does not describe the outer morphology appropriately. Nevertheless, a renaming of these sensilla would be inconvenient for the comprehension and comparability with already published data and would make it necessary to use different abbreviations.

Table 2:

Papilla, hair and knob sensilla are in some studies called campaniform, trichoid, and basiconic sensilla, respectively (Hartenstein 1988; Campos-Ortega and Hartenstein 1997; Green and Hartenstein 1997). Of course, these terms also have their validity and are therefore listed in Table 1. However, since in our view these terms were in part structurally more difficult to assign and could also lead to confusion from a functional point of view, we prefer the former. Trichoid sensilla in the adult olfactory system of Drosophila express odorant receptor genes and thus exert a chemosensory function (Couto et al. 2005; Miller and Carlson 2010; van Goes Naters and Carlson 2007). In contrast, any potential larval trichoid sensilla are likely to have only a mechanosensory function due to their ultrastructure.

In general, most external sensilla are arranged along the vertical (dorso-ventral) axis close to the middle of each segment (Figure 1B). In accordance with previous studies (Lohs-Schardin et al. 1979; Dambly-Chaudière and Ghysen 1986; Green and Hartenstein 1997), we found a stereotypic and fixed pattern of these sensilla. This structural consistency allowed us to generate a spatial map defining each sensillum and sense organ on the larval body by their precise position (Figure 1B). Furthermore, we were able to find the associated sensory and accessory (support) cells (Figure 1C). The spatial pattern and abundance of types of sensilla differs between segments. We observed three basic patterns. The sequence of sensilla is always similar for and different between T1, T2-T3, and A1-A7. The head and the last abdominal segments represent unique cases with specialized sense organs like the terminal sensory cones (t) (Singh and Singh 1984; Dambly-Chaudière and Ghysen 1986).

Papilla and hair sensilla are prevalent on all mapped segments whereas the keilin’s organs (KO), the knob sensilla, the double-hair and the sensory cones are not. The KO is exclusively found on thoracic segments; the double-hair is only found at abdominal segments. Knob sensilla are found on the thoracic segments, but also in the TO of the head and in the terminal sensory cones, although the number of neurons innervating these knob sensilla differs from one to three (Figure 1C, 13, 16). Papilla sensilla are most abundant in terms of number. We counted six papilla sensilla on the head, ten papilla sensilla on T1, seven on T2-T3, and nine on A1-A7 (Figure 1B and Figure 10). Please note that the numbers are given per hemisegment. We found three hair sensilla on T1-T3, and two on A1-A7 (Figure 1B and Figure 11). Two knob sensilla are restricted to the TO (Figure 16G) and T1 - T3 (Figure 1B and Figure 13). One KO is situated on the ventral side of each hemisegment of T1 -T3 (Figure 1B and Figure 14). One double hair can be found on A1-A7 (Figure 1B and Figure 12). Furthermore, one papilla, four hair and five knob sensilla can be found in segment A8/A9, but as mentioned before, they are organized in specialized terminal sense organs (Figure 1B, 15, 16, S3). In addition, we find one unpaired anal papilla and the spiracle sense organ on the last segment. This sensilla configuration is based on the single ssTEM and various SEM scans that we compared with published data (Lohs-Schardin et al. 1979; Dambly-Chaudière and Ghysen 1986; Green and Hartenstein 1997).

In the following, sensilla and sense organs of the larval head, thoracic and abdominal segments are classified based on their external and internal morphology. Also, developmental aspects are addressed by comparison of first (L1) and third instar (L3) larvae.

Description of individual external sensory organs of the Drosophila larva

DORSAL ORGAN

Despite previous studies on the ultrastructure of the DO (Chu-Wang and Axtell 1971; Singh and Singh 1984), many aspects of its structural organization, like the peripheral sensilla and the corresponding accessory cells (Acs), remained unclear. The 3D EM volumes recorded in the present work allow to complement and clarify the knowledge on the anatomy of the DO (Figure 2-7, Table 1). Figures 2-7 provide a comprehensive description of the organization of the entire organ. Therefore, by means of targeted FIB-SEM and ssTEM, we obtained continuous image stacks of the DO covering the distance from its outer sensory parts to the region of the DO ganglion (DOG), where the cell bodies of the sensory neurons are located. It enabled us to trace dendrites from their tips at the cuticle surface to the neuron’s cell body in the DOG and to generate a 3D reconstruction visualizing the spatial organization of the neuronal components of the DO (Figure 2A, C). Furthermore, accessory support cells were identified and described in detail (Figure 7). Most sensilla display a repertoire of these cells in a highly stereotyped fashion: a thecogen, a trichogen and a tormogen cell. Due to their role in formation of the sensillum, they also have been termed sheath, shaft and socket cell, respectively (Prelic et al. 2021).

The DO is the primary olfactory organ of the larva. Its central structure is the dome (Figure 2B-E), which comprises seven olfactory sensilla composing three dendrites each (Figure 3B-F), as reported in the literature (Chu-Wang and Axtell 1971; Singh and Singh 1984). Olfactory dendrites branch multifold in the dome-region of the DO and connect to pore tubules spanning the outer cuticle, each leading to one of the tiny pores perforating the dome (Figure 2B, D, E). Apart from these tiny pores, the dome is interspersed with seven larger pores arranged in a circle (Figure 2B, D). In contrast to L3, these pores are missing in L1 larvae (Figure 2E), which suggests that these are molting pores formed during ecdysis, when olfactory sensory neurons pass through the developing cuticle of the new DO. The seven olfactory triplets are housed in a common outer sensillum lymph space, which is built by the ACs (Figure 3B, C, E and Figure 7). In the outer lymph space, the triplets are enclosed by a dendritic sheath (Figure 3G), which is most likely segregated by the thecogen cell (Figure 7). Further distal, the three olfactory dendrites are bathed in a small inner sensillum lymph cavity enclosed by the thecogen cell (Figure 3D, F), also called the perineuronal lumen (Prelic et al. 2021). Inside the cavity lays the ciliary constriction (cc), which marks the transition from outer dendritic segment (ods) to inner dendritic segment (ids) (Figure 3G).

Figure 3:

All cell bodies of DO sensory neurons lie in the dorsal organ ganglion (DOG), in which the soma of olfactory receptor neurons (ORNs) are located more to the center compared to those of the peripheral sensilla (Figure 2A, C). The cell bodies of neurons innervating individual peripheral sensilla lie adjacent (Figure 2A, C). Cell bodies of olfactory neurons appear elliptically shaped and are relatively large compared to peripheral neurons’ cell bodies that have a roundish appearance and are smaller in diameter (Figure 2F, G). Also located within the DOG are additional cell bodies of the sensory cells of the papillum sensillum of the dorsolateral group of the TO (P_do), seven additional non-sensory cells of unknown function but most likely glial cells and of one additional papilla sensillum (p₆) in close proximity to the DO. Their cell bodies lie in the proximal end of the DOG from where they wrap around the whole ganglion.

The six peripheral sensilla lie in the rim between the dome and the surrounding cuticle ridge. They are not visible by external investigation (Figure 2B and Figure 3B, C). The outer cuticle part of all six peripheral sensilla appears similar in its structure. Its organization is very simple, consisting of a small cuticle bulge with a tiny pore in its center (Figure 3C). The pore is first surrounded by a short cuticle channel that leads to the dendrites (Figure 3C). Remarkably, the pores are absent in L1 larvae (Figure 3B) and are therefore most-likely molting pores. Internally, however, peripheral sensilla have different structural properties (Figure 4 and Figure 5).

Figure 4:

Figure 5:

The spatial arrangement of the peripheral sensilla was consistent in all analyzed samples. This stereotypical pattern allowed us to number the peripheral sensilla DO_p1 to DO_p6 (dorsal organ peripheral sensilla 1 to 6). DO_p1 is the posterior most sensillum. The six DO_p were numbered from 1 to 6 in a clockwise direction in the left and anti-clockwise direction in the right hemisphere when seen from above (see supplement Figure S2).

Within these six sensilla we find three structurally similar types (type 1: DO_p1; type 2: DO_p2 and DO_p4; type 3: DO_p3, DO_p5, and DO_p6; see Table 1). DO_p1 is characterized by a single dendrite with a tubular body that terminates at the base of the epicuticle (Figure 4B, C). DO_p2 and DO_p4 can be identified by the presence of two dendrites, one (DO_p2A and DO_p4A) ends with a tubular body, the other one (DO_p2B and DO_p4B) not (Figure 4D-G). Both dendrites are enclosed by a common dendritic sheath. Similar to the olfactory neurons, DO_p1/2/4 dendrites are bathed in a small inner sensillum lymph space which is enclosed by the thecogen cell (Figure 4C, E, G). DO_p3, DO_p5 and DO_p6 share a similar structural organization and house two dendrites each (Figure 5F, G). DO_p3A, DO_p5A and DO_p6A end below the epicuticle and are enclosed by a dendritic sheath (Figure 5B, C). DO_p3B, DO_p5B and DO_p6B end inside the inner sensillum lymph space.

The dendrites of DO_p3, DO_p5 and DO_p6 form the ciliary constriction inside an inner sensillum lymph cavity, like other DO sensilla. DO_p3/5/6-A have a small dendritic outer and a larger dendritic inner segment like canonical receptor neurons. In contrast, DO_p3/5/6-B are lamellated and form membrane staples, called dendritic bulbs (db), inside the lumen (Figure 5D-G). In L1 larvae, their dendrite terminates clearly inside the lumen, in L3 larvae, it might extend further into the dendritic sheath. Besides the dendritic bulb, the dendritic inner segment is heavily swollen after the ciliary constriction, forming another bulb-like structure (Figure 5E-G). This fits with confocal image data, where the dendritic bulb seems to be divided (Ni et al. 2016). In the DOG, cell bodies of both neurons lie adjacent (Figure 2A, C).

In addition to the peripheral sensilla DO_p1-6, we find one papilla sensillum, p₆, in close proximity to the DO, whose sensory cell body (des_1B) lies in the DOG. The structure of papilla sensilla will be presented in a separate results section (see Figure 10).

It is difficult to compare our findings to Singh and Singh (1984), because their description of peripheral sensilla is vague. In Musca, Chu-Wang and Axtell (1971) classified the peripheral sensilla into four types: contact chemoreceptor, unclassified receptor, lateral pore receptor and scolopidium-like receptor. Based on numbers of dendrites and other structural properties, DO_p4 and DO_p2 might correspond to the unclassified receptor and DO_p5 and DO_p3 to the contact chemosensory receptor. DO_p1 most likely corresponds to the scolopidium-like receptor, even though the classification is misleading, as scolopales are associated with chordotonal organs, which have no connection to the surface. From today’s perspective it would rather be classified as a campaniform or papilla sensillum, and misclassification might be due to interpretation of dendritic sheaths as scolopales in campaniform sensilla.

The seven olfactory sensilla composing the dome of the DO share a similar structural organization (Figure 3B-G). Therefore, discrimination between them, like done for the peripheral sensilla, was not feasible. At the base of the dome, olfactory dendrites form seven, tightly packed clusters of three dendrites (Figure 3B, C). These olfactory triplets lie in a circle immersed in common sensillum lymph space, which is filled with a substance of heterogeneous electron-density (Figure 3B, C). The peripheral sensilla are arranged as a circle around the olfactory triplets. They lie outside of the sensillum lymph space (Figure 3B, C, E). Dendrites of all DO sensilla are enclosed by ACs further proximal of the dome base (Figure 7). The olfactory sheath cells form the common olfactory receptor lymph space. Further proximal, the olfactory triplets are bathed in electron-dense inner sensillum lymph inside their individual lymph cavity (Figure 3D and F, Figure 7). In comparison to dendrites of the peripheral sensilla, olfactory dendrites appear quite large in diameter, whereas DO_p3, DO_p5 and DO_p6 appear very tiny, especially in L3 larvae (Figure 3C, E).

The dendrites of DO sensilla are separated into an inner and an outer dendritic segment by a ciliary constriction inside the inner sensillum lymph cavity (Figure 3G; 4C, E, G; 5E-G). Dendritic inner segments of olfactory and peripheral sensilla are considerable larger in diameter than the outer segments, which is the typical structure of insect sensilla dendrites (Keil 1997). A dendritic sheath surrounds the dendritic outer segments (e.g. Figure 3G; 4B, D, E; 5B, C) and disappears inside the lumen (Figure 3G). Dendrites of one sensillum seem to stay together after leaving their common lymph space and their cells bodies lie adjacent inside the DOG (Figure 2A, C).

In total, we find 43 cell bodies in the DOG, 36 of them being of sensory and seven being of non-sensory origin. The sensory cells in the DOG include the cell bodies of the 21 olfactory neurons, the 11 peripheral neurons, the neuron of the peripheral papilla sensillum and the three neurons of the papillum P_do of the dorso-lateral group of the TO (Rist and Thum 2017). The seven non-sensory cells are of unknown origin, but most-likely peripheral glial cells (see discussion).

Within the DO, we also find the (accessory) cells that build the structure of the organ. We find 60 cells in total, 28 (7×4) being associated with the olfactory sensilla, 24 (6×4) being associated with the six peripheral sensilla and one being associated with the papilla sensillum p₆. The seven remaining cells are of unknown origin, but might also be peripheral glial cells. The remaining ACs of p₆ and of P_do lay outside the DO.

CHORDOTONAL ORGAN IN CLOSE PROXIMITY TO DORSAL ORGAN

In close proximity to the DO we find a single-innervated (monodynal) chordotonal organ (ChO) (Figure 6A, B). It lays diagonally to the DO dome in between the non-sensory cells that build up the DO (Figure 6C, D). The bipolar sensory cell ends with a ciliary structure of type 1 (Yack 2004) (Figure 6E). The cilium is growing out of the dendritic inner segment and is surrounded by the prominent scolopale, which is made up of the scolopale rods (Figure 6D, E). The rods are segregated by the scolopale cell. The end of the cilium is inserted into a cap, which is ensheathed by the cap cell (Figure 6 D). The cilium is bathed in a sensillum lymph within the scolopale (Figure 6 E) and exhibits a ciliary dilation (Figure 6D). The inner dendritic segment contains a very noticeable striated ciliary rootlet (Figure 6F), which originates from the basal body towards the cell body. We find these these DO associated ChO (doChO) not only in L1, but also in L3 larvae (Figure 6G, white arrow).

Figure 6:

Figure 7:

VENTRL ORGAN

The ventral organ is located ventral of the terminal organ and lateral of the mouth hooks. The sensilla of the VO are located in a cuticle invagination, hidden by a row of cirri (Figure 8B). We identified four sensilla in the VO by targeted FIB-SEM and ssTEM investigation of internal and external morphology (Figure 8A-I, Table 1). Three of them are innervated by a single neuron and one by two neurons (Figure 8F-I). This number is in accordance with previous findings in other dipteran, cyclorrhaphan larvae (Chu-Wang and Axtell 1972b; Honda and Ishikawa 1987) but contradicts a study on Drosophila larvae reporting that five sensilla belong to the VO (Singh and Singh 1984).

Figure 8:

We name these sensilla VO₁-VO₄. Starting from the medial most sensillum VO₁ we number the four sensilla in a clockwise direction in the left and an anti-clockwise direction in the right hemisphere when seen from the front. VO₁ forms a shallow dome centering a tiny pore (Figure 8B). VO₁ is innervated by one dendrite that terminates with a tubular body at the base of the pore and is encased by a dendritic sheath (Figure 8C, F, G). The microtubules can be clearly distinguished and are evenly distributed in the tubular body area (Figure 8C’’). The VO₁ sensillum was also termed plate sensillum in earlier literature (Honda and Ishikawa 1987). VO₂, VO₃ and VO₄ lie in pits (Figure 8). In L3 larvae, we could observe terminal pores, which are absent in L1 larvae and thus are likely molting pores. VO₃ and VO₄, like VO₁, resemble papilla sensilla. All are innervated by only one dendrite, which composes a tubular body at its tip at the base of the pore openings or the epicuticle respectively (8 A, C, F, G). The tubular body of VO₃ consists of more densely packed microtubules which are not distinguishable from each other (Figure 8C’). VO₄ displays a unique type of tubular body, which can only be found in this sensillum. It is rather large compared to other tubular bodies, and the dendrite intermingles with electron dense material of unknown origin in this area (Figure 8C, D). In contrast to the other three sensilla, VO₂ is innervated by two dendrites (Figure 8H, I). In contrast to the dendrites innervating the other sensilla of the VO, the dendrites of VO₂ are surrounded distally by a cuticle tube (Figure 8A, C), lack a tubular body and branch multifold distally of the ciliary constriction (Figure 8E). The terminal pore is also present in L1 larvae. Each of the VO sensilla possess their individual set of three ACs (Figure 8 F-I), but the thecogen and trichogen cell of VO₂ appear substantially more electron-lucent (Figure 8H, asterisks), with mitochondria of altered structure (Figure 8H, white arrowhead) compared with ordinary ones (Figure 8H, black arrowhead).

LABIAL ORGAN

Located on the ventral side of the ventral lip (labrum) lays the labial organ (LO) (Figure 9A, B). In the present study, we find two sensilla associated with the LO in accordance with Kankel et al. (1980) (Figure 9C, I). In contrast, Singh and Singh (1980) describe three sensilla based on examination of internal ultrastructure. We name the two identified sensilla LO₁ and LO₂ (Figure 9C). LO₁ forms a cavity with a pore in the center and sits on a small, shallow socket (Figure 9C). The pore is absent in L1 larvae and thus likely a molting pore. LO₁ is innervated by one neuron that composes a tubular body on the tip of its dendrite, which terminates below the pore opening. The dendrite is surrounded by a dendritic sheath (Figure 9E, F). LO₂ forms a knob-like cuticle shaft protruding from the cuticle. The knob’s cuticle has a rough texture and appears more electron-lucent than the cuticle of the surrounding body wall (Figure 9C). Internally, the knob is filled with an electron-dense material (Figure 9E). Like LO₁, LO₂ is innervated by one dendrite composing a tubular body, which terminates at the base of the knob where a molting pore is present in L3 larvae and absent in L1 larvae (Figure 9D-F). The two labial organ sensilla both have their own set of ACs, a thecogen cells which forms the dendritic sheath, a trichogen cell which forms the shape of the sensillum and a tormogen cell which forms the sensillar socket. We find two more cells in the LO, most likely non-apoptotic glial cells originating from the sensory organ precursor cell (SOP) or its secondary precursor cell (pIIb), respectively (Fichelson and Gho 2003).

Figure 9:

Sensilla of thoracic and abdominal segments

PAPILLA SENSILLA

The papilla sensillum is most similar to the canonic type of the campaniform sensillum. Papilla sensilla form a shallow depression in the cuticle (Figure 10A, B) with a pore in its center in L3 larvae (Figure 10B, D). In L1 larvae, the pore is absent and therefore a molting pore (Figure 10C, E). Papilla sensilla are innervated by one dendrite which terminates with a tubular body below the pore or the epicuticle, respectively (Figure 10C-E). The tubular bodies show no organized distribution of microtubules which are difficult to distinguish as they occur in densely packed clusters (Figure 10E). The tubular body of the p₆ sensillum in the abdominal segments is of a similar shape, although there exists another sensory neuron without a tubular body in a shared sensillum space (Figure 10G). The dendrites of abdominal sensilla p₅, so called slit papilla, and the p_y papilla of the first thoracic segment don’t show the typical dendritic swelling at tip and the tubular body appears to be more delicate (Figure 10G’, G’’). In contrast, the tip of the thoracic p_x neuron is thickened and the tubular body is quite noticeable, as the whole inner area is packed with electron-dense material (Figure 10G’’’). For all types, the tip of the dendrite is anchored in the endocuticle by a socket septum (Figure 10C-E). The dendrites are enclosed by a dendritic sheath which is most likely segregated by the thecogen cell (Figure 10C-E). The thecogen cell also forms an inner sensillum lymph cavity at the transition from outer to inner dendritic segment at the level of the ciliary constriction (Figure 10C, H, I). Furthermore, the sensillum is enveloped by thecogen and tormogen cell (Figure 10H, I). In some exceptional cases, we find a short hair-like structure protrude from papilla sensilla or positions where we would expect papilla sensilla (Figure 10F).

Figure 10:

HAIR SENSILLA

A hair sensillum (Figure 11A) is most similar to the canonic type of the trichoid sensillum. Hair sensilla comprise a round, hair-shaped shaft that sits in the center of a shallow cuticle depression (Figure 11B). In accordance with previous literature (Kankel 1980), we find that the shaft of hair sensilla varies greatly in size. It might be very short, reduced to a stump, or very long up to more than 15 µm (Figure 11B). The form of the hair-shaft might vary, too. We observed bifurcated shafts forming two branches (Figure 11B’). Further, we observed hair sensilla that come in a pair (‘double hair’, Figure 12). The double hair h₃/h₄ was exclusively found on abdominal segments. Because of the differences in external morphology of the hair-shaft, Kankel et al. (1980) classified hair sensilla into three different types, called type C, D and E (Table 2). However, we here find that the internal ultrastructure of hair sensilla (Figure 11C-I) is in general similar irrespective of length and shape of the shaft. The interior of the hair-shaft is electron-lucent, surrounded by an electron-dense sheath (Figure 11C, D). A pore, presumably a molting pore, can be found at the base of the shaft (Figure 11D) in L3 larvae. All hair sensilla are innervated by a single dendrite terminating at the base of the shaft composing a tubular body (Figure 11E, F). The tubular body of the hair sensilla is even more thickened than the canonical tubular body of papilla sensilla, and the microtubules are clearly visible and distinguishable from each other. A socket septum is clearly visible (Figure 11E, F). Apart from the difference in outer appearance and structure of the tubular body, the hair sensilla are quite similar to the papilla sensilla, with a typical set of ACs (Figure 11G - I) and a small inner sensillum lymph cavity at the transition from outer to inner dendritic segment at the level of the ciliary constriction (Figure 11E, H, I).

Figure 11:

DOUBLE HAIR SENSILLA

As mentioned previously, the abdominal hair sensilla h₃ and h₄ represent a special case of hair sensilla. It consists of two hairs of different sizes which are adjacent to each other sitting in one cuticle depression (Figure 12A, B). Usually, the h₄ sensillum structure is of the same type as a canonical hair sensillum, containing one sensory cell with a tubular body at the base of the hair (Figure 12D). In contrast, the outer hair of h₃ is comparatively short and therefore called a bristle. The key aspects are similar to the abdominal papilla p₆, with two sensory cells (Figure 12 E, F), one containing a tubular body (which rather exhibits the structural properties of papilla than hair sensilla) (Figure 12C, D). Both h₃ and h₄ possess their own set of ACs and are therefore individual sensilla (Figure 12H, I), although in one exceptional case, they shared one lymph space (Figure 12G). At the level of the ciliary constriction, the thecogen cell forms an inner sensillum lymph cavity (Figure 12E, F).

Figure 12:

KNOB SENSILLA

The knob sensillum is most similar to the canonic type of the basiconic sensillum. Knob sensilla are present on thoracic segments (Figure 13A) and on the sensory cones of the last abdominal segment (Figure 16A), but similar structures can also be found at the terminal organ (Figure 16G) (Rist and Thum 2017). Knob sensilla have been described under vaious names such as koelbchen, knob-in-pits, hair-type B, black sensory organs, black dots, sensory papillae or dorsal/ventral pits (Hertweck 1931; Lewis 1978; Lohs-Schardin et al. 1979; Kankel 1980; Singh and Singh 1984; Campos-Ortega and Hartenstein 1985; Sato and Denell 1985; Dambly-Chaudière and Ghysen 1986; Hartenstein 1988; Campos-Ortega and Hartenstein 1997) (Table 2).

Figure 13:

The knob sensilla of the thoracic segments display a common external morphology, a knob-shaped cuticle shaft sunken into a round cuticle cavity (Figure 13B). The cuticle shaft slightly protrudes from the cavity. Investigation of internal ultrastructure by targeted FIB-SEM and ssTEM reveals that three dendrites are associated with one thoracic knob sensillum (Figure 13C-I). One dendrite innervates the shaft proceeding to its tip (Figure 13C). The other two dendrites end at the base of the knob, one with a tubular body (Figure 13C-E). All dendrites are surrounded individually by one dendritic sheath (Figure 13C-E), which arises from the common thecogen cell. Each sheath is connected to a pore at the base of the knob in L3 larvae (Figure 13D). This pore is not present in L1 larvae (Figure 13 C, E) and is therefore most likely a molting pore. The dendritic sheath of the three dendrites disappears at the epidermal layer. Here, dendrites are surrounded by three common sheath cells (Figure 13H, I). In T1, we find one dorsal knob sensillum (dk) and one ventral knob sensillum (vk), whereas in T2- T3, we find one lateral knob sensillum (lk) and one ventral knob sensillum (Figure 1B and Figure 13A). All thoracic knob sensilla were examined and no difference in structural organization was recognized. In contrast, the knob sensilla of the sensory cones are inconsistent and house either two (Figure 16D, E) or three (Figure 16F) neurons, whereas the knob sensilla of the TO only contain one sensory cell (Figure 16G) (Rist and Thum 2017), which dendrite protrudes into the shaft. Knob sensilla of the last abdominal segment are discussed in a separate results section, as they are organized in specialized terminal sense organs.

KEILIN’S ORGAN

The KO is exclusively located on the ventral side of the thoracic segments T1-T3 (Figure 14A). The KO is relatively easy to identify and was described for larvae of several dipteran species (Keilin 1915; Lakes and Pollack 1990; Lakes-Harlan et al. 1991). The KO of Drosophila larvae was so far described as consisting of three hairs (Kankel 1980). In L1 larvae, we find that two papilla-like sensilla are associated with the organ in addition to three hair sensilla (Figure 14 A, E, G). One of the papilla-like sensilla is degenerating during the L1 stage (Figure 14E-F). Signs of degeneration are the comparatively small diameter of the dendrite (Figure 14E) and a poor axonal development, like the absence of growth cones and axon branching (not shown). This finding is congruent with a study of the KO of the larvae of Phormia. (Lakes-Harlan et al. 1991) and provides an explanation for studies that find five sensory neurons innervating the KO of Drosophila larvae (Dambly-Chaudière and Ghysen 1986; Campos-Ortega and Hartenstein 1997). In L3 larvae, the (surviving) papilla sensillum is externally recognizable by a tiny pore in the cuticle (Figure 14B). Also, a pore is found on the base of each of the hairs (Figure 14D). Because these pores are not found in L1 larvae, they are most likely molting pores. Investigation of the internal ultrastructure reveals that the three hair and the two papilla sensilla are associated with a single dendrite. All dendrites terminate with a tubular body below the cuticle; this organization represent the standard types of hair or papilla sensilla (Figure 14E, F). Also, all KO sensilla possess their own set of enveloping ACs (Figure 14H, I).

Figure 14:

Sensilla of the anal division

TERMINAL SENSORY CONES

The terminal sense organs or sensory cones are located at the fused terminal segments (Figure 1A, Figure 15A, Figure 16A). According to the literature (Whittle et al. 1986; Denell and Frederick 1983; Dambly-Chaudière and Ghysen 1986; Jürgens 1987), we could identify seven distinct sensory cones t₁ – t₇. They either house a knob sensillum (t₃, t₅, t₇), a hair sensillum (t₄, t₆), both (t₂) or both plus an additional papilla sensillum at the base of the cone (t₁) (Figure S3). These results are consistent with previous findings in L1 larvae (Dambly-Chaudière and Ghysen 1986; Jürgens 1987), although the number of corresponding neurons is slightly different. We found that knob sensilla of t₁, t₅ and t₇ are only innervated by two sensory neurons (Figure 16E) in contrast to three sensory neurons that were reported in the embryo (Dambly-Chaudière and Ghysen 1986) and that we find in t2 and t3 (Figure 16C). These knob sensilla are still innervated by one neuron protruding into the shaft and one neuron containing a tubular body that sits at the base of the shaft. A third neuron without a tubular body is absent (Figure 16E). The hair sensilla associated with the sensory cones are innervated by one sensory cell with a tubular body on its tip (Figure 15B) and exhibit similar features as the canonic hair sensilla of the body wall. The only papilla sensillum associated with the sensory cones sits at the base of the t1 and shares the same structure as canonic papilla sensilla. The t₁ lays in close proximity to the anal plate towards the caudal end. t2 is situated dorso-caudal of t₁, halfway in between the anal plate and the posterior spiracle. t₃ lays anterodorsally of t₂ and the cone of t₄ lays anterodorsally of t₃. The cone of t₅ is located dorsal of t₄ wheres t₆ is located closer to the dorsal midline than t₅. The cone of t₇ is located at the dorsal midline and at the base of the posterior spiracle. All knob sensilla of the terminal sensory cones show a similar external structure as thoracic knob sensilla, but their shafts protrude far out of the cavity in L1 larvae (Figure 16B, C; Figure S3). In L3 larvae, the sensory cones further change in appearance, and the knob and hair sensilla are sunken into the full-grown cone and surrounded by broad based apposing leaflets (Kuhn et al. 1992). All sensilla associated with the terminal sensory cones display a typical sensilla configuration: their dendrites are bathed in an inner sensillum lymph and ensheathed by a typical set of support cells (Figure 15 C, D and Figure 16F, H, I). Besides the prominent sensory cones, there are some less noticeable external sensilla. The spiracle sense organ (sp) consists of four sensilla (sp_A-D) which are located in close proximity of the spiracular openings at the base of the spiracular hairs (Figure 17A-D). They are all single-innervated, contain a tubular body (Figure 17D) and display a canonic set of support cells (Figure 17 E, F). The spiracular glands lay in close proximity to the these sensilla (Figure 17E, F). One papilla sensillum can be found at the anal plates. It is located at the anal opening and called ventral anal papilla sensillum (vas) in the literature (Campos-Ortega and Hartenstein 1997). In contrast to Hartenstein’s findings, we find it to be only single innervated. Notably, one unpaired anal cone or anal tuft (Lohs-Schardin et al. 1979; Denell and Frederick 1983) is located in close proximity, that was mistaken as a sensory cone before. We could not find any sensory cell associated with this structure, although vas lies in close proximity (Figure 1B).

Figure 15:

Overall, we obtained a complete picture of all larval external sensory structures and their associated sensory neurons. In the following, we discuss their putative sensory mechanism and function according to their internal morphology and ultrastructure and classify the different sensilla and their associated neurons.

Discussion

We present a comprehensive anatomical analysis of external sensory organs of the head, thoracic and abdominal segments of the Drosophila larva. The application of optimized FIB-SEM and ssTEM allowed us to image and reconstruct small sensilla as well as large and complex sensory organs. Our experimental FIB-SEM approach can easily be adapted to study the morphology of external sensilla and other external structures of further insect species. This methodology is a prerequisite to investigate the different characteristics of insect peripheral nervous systems in a comparative way. With respect to Drosophila larvae, our work provides fundamental anatomical insights necessary to further decipher the relation between the structure and function of sensory organs. Certainly, a future key step would be to establish transgenic fly lines that label the sensory neurons of the sensilla allowing the study of their functionality through molecular, behavioral and physiological experiments. This might be possible through a variety of genetic driver lines from different Gal4, LexA, and Q system collections (Lewis 1978; Brand and Perrimon 1993; Lai and Lee 2006; Pfeiffer et al. 2010; Li et al. 2014; Gohl et al. 2011; Tirian and Dickson 2017). In addition, there are even more specific lines based on multiple overlapping promoter expression patterns such as the split-Gal4 system (Luan et al. 2006). Unfortunately, these lines have mostly been studied only with respect to their expression in the central nervous system, and anatomical studies of the peripheral nervous system are limited. It will be equally important to connect the described peripheral sensory system of the larva with the already existing brain connectome in order to better understand the processing of external environmental information and its relevance for behavior (Hückesfeld et al. 2021; Miroschnikow et al. 2018; Eichler et al. 2017; Eschbach et al. 2020; Ohyama et al. 2015). Such linkage would be important, for example, to analyze whether mechanosensory information from different body segments converges in a the olfactory system (Berck et al. 2016; Ramaekers et al. 2005). Similarly, it would allow us to understand for the first time how taste information from different external sensory organs is interconnected with the internal pharyngeal and enteric nervous systems.

The standard configuration of the larval body wall covers 541 sensory cells in total associated with the external sensilla. 363 are probably mechanosensory according to their internal structure as they possess a tubular body. The 363 mechanosensory cells can further be subdivided into 201 neurons belonging to papilla sensilla, 100 to hair sensilla and 22 to knob sensilla. The remaining 40 cells can be found in the major head organs and in the spiracle sense organ. We find 42 olfactory and 12 thermosensory neurons in the DOs and 50 gustatory neurons of which 46 cells are distributed in the TOs and 4 in the VOs. We find additional 26 sensory neurons in knob sensilla that are probably important for oxygen and carbon dioxide sensing. Their dendrites protrude into the sensillum shaft where the dendritic sheath is perforated and therefore susceptible to chemical compounds that are able to diffuse into the cuticle. We also find 48 sensory cells whose ultrastructure was insufficient to make well-grounded predictions about their sensory mechanism, as they are lacking noticeable structural components. These cells are distributed evenly throughout the larval body. We find two in each of the DOs, one in each thoracic knob sensillum, one each in the abdominal papilla sensilla p₆, one in the abdominal hair sensilla h₃ and one each in the knob sensilla of the terminal sensory cones t₂ and t₃. Apart from these sensory cells, there are more to particular brain region and forms a stable, spatially conserved map of inputs - similar to be found in the larval body, like in chordotonal organs, multidendritic neurons or in pharyngeal sense organs. These were not regarded in this study but will be the subject for future studies.

Organization of larval sensilla types

OMENCLATURE, VARIABILITY AND ULTRASTRUCTURE

As mentioned before, various names were given to the sensilla not only of immature insects but also larval Drosophila. Here, we aimed to list all names given to them (Table1 and 2) and to establish a standard nomenclature for Drosophila larval sensilla, using our detailed and comparative approach as a basis for classification of all external sensilla. We identified three major types of sensilla which are most abundant on the larval body wall: papilla sensilla, hair sensilla and knob sensilla. They occur either as solitary sensilla or are integrated in sensory organs or compound sensilla like the terminal organ, the keilin’s organ or the terminal sensory cones. Also, they exhibit slight differences in terms of associated sensory cells or external appearance. Besides these most abundant types, we found specialized (olfactory and gustatory) sensilla, mainly in the organs of the head region.

Taking advantage of the whole larval volume, we were able to identify and name all neurons associated with the external sense and chordotonal organs (Figure 1C). Again, we used the nomenclature of Dambly-Chaudière and Ghysen (external sense) and Campos-Ortega and Hartenstein (chordotonals), but made slight modifications to incorporate deviating cell numbers and to create a consistent nomenclature that can be used in databases. These changes are: (i) for all external sensilla, the associated neurons are named (2), v’’entral (3)), type of sensillum (external sensilla/sense organs or chordotonal organs), number of associated neurons (1, 2, 3,…) and order of appearance from ventral to dorsal (A, B, C, …); (ii) for the fused head and abdominal segments, we also used this pattern for simplification and clarity, even though we thereby ignore the segmental borders; (iii) to avoid confusion, we did not assign names that were used for other neurons again, for example, the neurons of abdominal hair sensillum h₃ were named des_A before, as Dambly-Chaudière and Ghysen (1986) only could identify one neuron in their camera lucida drawings in the embryo. On the contrary, we did always see two neurons innervating h₃, and therefore called these des_2A. The remaining dorsal external sense neurons did not change in naming, even though des_1A is not assigned. A full list for the external sense and chordotonal neurons is given in Table 3. In the following, we discuss different criteria used to assign putative functions to the identified sensilla and sensory cells.

TUBULAR BODIES

Within the larval sensilla, we find a variation in shape and size of tubular bodies, a well-known structure regarded to mediate mechanical sensation in insect sensilla sensory neurons (Thurm 1964; Keil 1997). These variations likely correlate with different mechanisms of mechanotransduction. For example, the tubular body of hair sensilla is shaped in a way to transduce directional force applied to the hair (see Figure S4). When the hair is deflected, pressure is either applied or in the following pattern: position within the segment (dorsal, lateral, ventral (1), v’entral released from the dendrite tip via the cuticular structures (like a lever). In papilla sensilla, the force is rather applied unidirectional, either through deformation of the cuticle through external touch or movement of the body wall during motion. Abdominal p₅ papilla are contained in vertical slits (slit papilla), and the tubular body sits beneath an elongated, cuticular mound in the center of the slit (Figure 10G’, G’’). Putatively, the tubular body is deformed by compression of the slit during movement. In contrast to the other tubular bodies (and especially to the delicate tubular bodies of p₅ and py sensilla), the tubular body of px at the first thoracic segment appears to be very dense (Figure 10G’’’). One may assume that a dense tubular body is also more rigid and, because it acts as an abutment for the spring like proteins that work as mechanotransducers, therefore makes the mechanotransduction apparatus more sensitive. However, more data on the physiology of these sensory neurons is necessary to provide evidence for such a structure-function hypothesis.

DEVELOPMENTAL ASPECTS

Morphological studies on the ultrastructure of the sensory neurons in Drosophila were either executed on Drosophila embryos (Campos-Ortega and Hartenstein 1997; Dambly-Chaudière and Ghysen 1986), first (Hartenstein 1988), second (Singh and Singh 1984) or third instar larvae (Dambly-Chaudière and Ghysen 1986). Over the years, these works led to contradictory assumptions about the number and location of the larval sense organs and their corresponding sensory cells. In addition, various functions were to some extent assigned to the sensilla. Our analysis, using L1 and L3 larvae, allowed for clarification of several ambiguities. Primarily, we could address the difference between functional pores and molting pores. Pores that were found in L3 larvae but absent in L1 larvae were considered to be traces of molting, and therefore not necessary for the sensory function of these sensilla. In most cases this was true for mechanosensitive sensilla that are widely distributed on the larval body wall. Given their pore-like appearance in L3, especially papilla sensilla were suspected to serve chemosensory function in the past. Additionally, the sensilla of the KO were assumed to serve different sensory function before benz (Hafez 1950; Benz 1956) (Benz 1956, Hafez 1960) and some of the peripheral sensilla of the DO were termed lateral pore receptor (Singh and Singh 1984; Chu-Wang and Axtell 1971) or classified as contact chemoreceptors (Chu-Wang and Axtell 1971). If pores are absent in sensilla of L1 larvae, they probably do not serve chemosensory function, although we cannot rule out this function completely given the lack of physiological experiments. On the other hand, pores that were found in both instars were considered to be functional pores, and therefore most likely associated with external gustatory sensilla. Pore-containing sensilla are limited and restricted to the VO₂ of the ventral organ (Figure 8C, E) and the pit and papilla sensilla of the terminal organ (data not shown, see results of (Rist and Thum 2017)).

Secondly, we could address the open questions regarding the number of sensilla and sensory neurons. For example, Singh and Singh (1984) reported five sensilla in the VO, but our results show only four sensilla in L1 as well as in L3 larvae. Furthermore, they reported five dendrites in VO₂. Because we obtained volumes of the VO including the sensory cell bodies, we could show that VO₂ is innervated by two sensory cells with one proximal dendrite each, which branch multifold when entering the pore channel. In the abdominal double hair, the h₃ sensillum houses two sensory cells instead of one reported before (Campos-Ortega and Hartenstein 1997; Dambly-Chaudière and Ghysen 1986). The keilin’s organ was reported to house either four (Lakes-Harlan et al. 1991) or five (Campos-Ortega and Hartenstein 1985; Campos-Ortega and Hartenstein 1997; Dambly-Chaudière and Ghysen 1986; Lakes-Harlan et al. 1991) sensilla previously. With our work, we could explain these differences: one papilla sensillum is still abundant in L1 larvae, but already shows signs of degeneration and is completely missing in L3 larvae (Figure 14F). This is the only example of age specific degeneration of sensilla between larval moltings that we could identify.

Hypothetical functions of larval head, thoracic, and abdominal sensilla

DORSAL ORGAN

In Drosophila, the six peripheral sensilla of the DO were hypothesized to serve different modalities: contact chemosensation (gustation), mechanosensation, thermosensation and hygrosensation (Chu-Wang and Axtell 1971; Python and Stocker 2002; Kwon et al. 2011; Klein et al. 2015; Ni et al. 2016; Hernandez-Nunez et al. 2021). Evidence derives from observed ultrastructure indicating certain functions (Chu-Wang and Axtell 1971), and more recently from anatomical, physiological and behavioral experiments (Kwon et al. 2011; Klein et al. 2015; Ni et al. 2016; Hernandez-Nunez et al. 2021). We conclude that DO_p3, DO_p5 and DO_p6 are candidates for thermosensory function, which was confirmed recently (Hernandez-Nunez et al. 2021). Typically, insect thermosensory sensilla are associated with neurons that terminate below the other neurons of the sensillum and form extensive lamellation (Steinbrecht 1984). DO_p3, DO_p5 and DO_p6 each house two neurons. One neuron ends right below the cuticle surface of the sensillum (Figure 5A, B) and the other becomes visible within the inner sensillum lymph cavity and forms a lamellated to act together as synchronous and opponent thermosensors (Hernandez-Nunez et al. 2021).

Contact chemosensory function was assumed for at least two peripheral DO sensilla based on structural characteristics in Musca larvae (Chu-Wang and Axtell 1971). Further evidence for gustatory function of the DO came from the report of Gr2a and Gr28a gene expression in DO neurons of Drosophila larvae (Colomb et al. 2007; Kwon et al. 2011). DO_p2 and DO_p4 each house two neurons of which one is likely to be mechanosensory. This function is indicated by the presence of a tubular body located in the tip of the neurons’ dendrite (Figure 4D, F). However, their sensilla structure differs from the gustatory sensilla found in the TO (Rist and Thum 2017). There, dendrites of one gustatory sensillum are surrounded by a common dendritic sheath, which forms the lymph-filled lumen of the sensillum. DO_p2 and DO_p4 are of simpler organization lacking an obvious sensillum lumen as each dendrite is enclosed by an individual dendritic sheath. A cuticle tube, present in all gustatory TO sensilla, is absent, too. Furthermore, a terminal pore is missing in L1 larvae, which arguments against a role in contact chemosensation. In dendritic bulb (Figure 5E, F). This remarkable structure might be similar to the ‘dendritic bulb’ described by recent studies using fluorescence microscopy of transgenic fly lines (Klein et al. 2015; Ni et al. 2016; Hernandez-Nunez et al. 2021). It was shown that the DO neurons associated with these structures trigger cool avoidance behavior in larvae (Klein et al. 2015). Presumably, this response is mediated by the Ir21a (Ni et al. 2016). In addition, two of these sensilla play a role in warmth sensing via the Ir68a. This receptor is expressed in the neurons without a dendritic bulb. Both neurons were shown adult Drosophila, Gr2a and Gr28a have been reported to be expressed not only in gustatory, but also in other sensilla of another, yet unknown function (Thorne and Amrein 2008). In addition, Gr28a was also reported to be expressed in the larval gut (Park and Kwon 2011), where gustation is not required. Therefore, we question a chemosensory function of DO_p2 and DO_p4. The structural organization of DO_p1 and papilla sensillum p₆ indicates mechanosensation due to the presence of a single neuron composing a tubular body at its dendritic tip (Figure 4B). Therefore, we conclude mainly mechanosensory function for p₆, DO_p1, DO_p2 and DO_p4, although DO_p2B and DO_p4B might be involved in a different sensory mechanism.

The DO has been reported to consist of 14 sheath, 14 shaft and 13 socket cells (Grillenzoni et al. 2007). Therefore, it has been concluded that the DO derives from 14 sensory organ precursor cells (SOP) and consists of 14 sensilla. With our data, we could confirm and even extend these assumptions. We hypothesize that the DO consist of 14 sensilla too, although the 14^th sensilla is not the P_do but the papilla sensillum p₆. Some of its ACs lie outside the DO, which is possibly the reason for the missing 14^th socket cell. P_do on the other hand, has its sensory cell bodies in the DOG, but all the ACs lie outside the DO. Furthermore, we could find in total 60 cells within the DO and 43 cells in the DOG. In one exceptional case, we observed an 8^th olfactory triplet in the DO, leading to an altered cell count of three more sensory cells in the DOG and four more cells inside the DO. Therefore, we hypothesize that the olfactory DO sensilla, if not all (DO sensilla), possess four ACs derived from the SOPs. If the fourth one is a 2nd tormogen cell, like in some adult olfactory sensilla (Shanbhag et al. 2000; Nava Gonzales et al. 2021), or a non-apoptotic glial cell, can’t be finally resolved with our data. In addition to the hypothesized 53 accessory cells (4 x 7 olfactory ACs, 4 x 6 peripheral ACs, 1 x p₆ glial cell) and 36 sensory cells (21 olfactory, 11 peripheral, 1 p₆ and 3 P_do neurons) we find seven cells of unknown origin in the DO and in the DOG respectively. They either also derive from SOPs or, more likely, are peripheral glial cells generated in the CNS that migrate towards their final destination (Hilchen et al. 2008).

VENTRAL ORGAN

We conclude that the VO is mainly a mechanosensory organ and its only gustatory sensillum might be VO₂. Contrary, a mainly gustatory function of the VO was assumed previously based on the presence of sensillum-terminal pores and the expression of gustatory receptor (Grs) genes in neurons of the VO (Chu-Wang and Axtell 1972b; Singh and Singh 1984; Python and Stocker 2002; Colomb et al. 2007). In fact, the terminal pores of most VO sensilla might be traces of molting, as they were absent in L1 larvae of onion and seed-corn flies (Honda and Ishikawa 1987), which have not undergone molting in this stage. Our results validate these findings, as there were no visible terminal pores (Figure 8C). VO sensilla VO₁, VO₃, and VO₄ display structural characteristics typical for mechanosensation (Keil 1997): innervation by a single dendrite terminating with a tubular body (Figure 4C). VO₂, in contrast, displays a morphology similar to the gustatory pit sensilla T₃ and T₄ of the TO (Rist and Thum 2017). Therefore, VO₂ might be the single sensillum of the VO serving contact chemosensation.

LABIAL ORGAN

We conclude that the main function of the LO lies in the detection of mechanical stimuli, too. The two single neurons of the LO sensilla each compose a tubular body at their endings (Figure 9E, F) indicating mechanosensory function. On the other hand, we cannot exclude additional sensory function, especially for the LO₂, which shows a noticeable outer appearance. It is possible that the club-shaped structure acts in hygrosensation by exerting pressure to the tubular body depending on humidity levels, acting as a hygromechanical transducer (Tichy and Kallina 2010).

PAPILLA SENSILLA

Papilla sensilla are the most abundant type of sensory structure throughout the larval sensory system (Figure 1B and Figure 10A). They occur in each segment including the anal division and the head capsule. We conclude that their main function is the detection of mechanical stimuli, given the fact that they all possess a tubular body at their dendrites tip (Figure 10C-G). Most of the papilla sensilla probably detect mainly deformation of the cuticle by motion or external pressure applied, although the direction of force detection might differ in some papilla like the abdominal slit papilla p₅. The same is true for the abdominal papilla sensilla p₆, but they house two sensory cells, v’es_2A, one of them lacking a tubular body and therefore possibly serving different sensory function.

HAIR SENSILLA

Hair sensilla are abundant on the thoracic and abdominal segments. Like papilla sensilla, their sensory neurons all end with a tubular body at the dendrites tip, indicating mechanosensation (Figure 10C, E, F). As discussed above, hair sensilla most likely transduce directional force applied to the distal end of the hair. In addition to mechanosensory function, another unknown function can be assumed for the abdominal hair sensilla h₃. It houses two sensory cells, des_2A, one of them with and one without a tubular body. The sensory function of these neurons is difficult to assume only by ultrastructural analysis, as they are lacking obvious structures like pores (gustatory), tubular bodies (mechanosensory) or dendritic bulbs or lamellated outer dendritic segments (hygro-and thermosensory).

KNOB SENSILLA

Hygro-and thermosensory function was assumed for Knob sensilla, (Hartenstein 1988), based on structural similarities with hygro-and thermosensory insect sensilla (Altner and Prillinger 1980; Altner et al. 1981; Steinbrecht 1984). Thereafter, knob-shaped sensilla sunken in a cuticle cavity associated with three dendrites typically hint towards a hygro-and or thermosensory function. One of the three neurons typically associated with hygro-and thermosensory sensilla displays peculiar structural characteristics as it might be branched or is highly lamellated in its tip that ends beneath the knob (Altner and Prillinger 1980; Altner et al. 1981; Steinbrecht 1984). Instead of lamellation, we find that one neuron of the thoracic knob sensilla contains a tubular body (Figure 13D, E). Knob sensilla organized in sensory cones also display one sensory cell containing a tubular body, irrespective if they are innervated by two or three neurons in total (Figure 16D, E). This result indicates a mechanosensory and not a hygro-and thermosensory function for these neurons. In the absence of physiological data, however, the sensory function of the knob sensilla remains unclear. A sensory function of the other two knob sensilla neurons in contact chemosensation is not inconceivable. The knob sensilla of the TO are similar in external morphology and are candidates of contact chemosensation due to the indicated expression of certain Grs in their sensory neurons (Rist and Thum 2017). In sensory neurons of thoracic segments, too, expression of Grs (Gr2a) was assumed based on GAL4 driver line analysis (Colomb et al. 2007). To which sensillum type these sensory neurons belong can now be determined based on the presently established morphological classification.

Additionally, knob sensilla of the larval thoracic segments were associated with two atypical soluble guanylyl cyclases (asGC), Gyc-89Da and Gyc-89Db, which were shown to sense oxygen and are important for response to hypoxia and in larval ecdysis (Vermehren-Schmaedick et al. 2010; Morton et al. 2008). Two neurons were found on each thoracic hemisegment which express both subunits, which fits with the number of thoracic knob sensilla. In addition, each subunit was found to be expressed in two different neurons innervating the TO (per side), which also fits the number of knob sensilla in the TO (Rist and Thum 2017). Furthermore, both subunits were co-expressed in neurons innervating the terminal sensory cones. It was stated that this is true for all cones (Morton et al. 2008). However, the assigned position seems to be slightly off. We hypothesize that these asGC were actually found in cones containing a knob sensillum (t₁, t₂, t₃, t₅, t₇). The sixth one, which also appears to be of different color in the confocal data, is most likely a tracheal dendrite (td) neuron, which we found to occur in this region (not shown). This interpretation is supported by findings, that these asGC subunits were also individually expressed in two td neurons in A1 and A2 (per hemisegment) (Morton et al. 2008). Because the knob sensilla of the TO are only innervated by one neuron which protrudes into the knob structure, we conclude that these neurons in general (for all knob sensilla) are the ones expressing the asGC. Also, within the knob, the dendrites are swollen and not encased by a dendritic sheath, which makes them the most-likely candidates for sensing external cues like oxygen (Figure 13C, Figure 16C, G). Therefore, we conclude that knob sensilla mainly serve chemosensory function (oxygen perception) and mechanosensory function (latter is not true for TO-associated knob sensilla). K₂ in the TO was also associated with CO₂ perception, as its single neuron expresses Gr21a and Gr63a (Rist and Thum 2017), which were shown to mediate CO₂ responses (Kwon et al. 2007, 2011; Jones et al. 2007; Faucher et al. 2006). In contrast to other larval knob sensilla, the dendrite of K₂ is lamellated at its tip, which could also indicate a hygrosensory mechanism (Altner and Prillinger 1980; Rist and Thum 2017). The function of the third neuron innervating the thoracic and some of the knob sensilla of the anal division is not known and will have to be determined in the future.

Figure 16:

KEILIN’S ORGAN

The KO of Drosophila larvae was assumed to serve hygro-and thermosensory function (Hafez 1950; Benz 1956). However, electrophysiological recordings of the KO of Phormia larvae indicate mechanosensory function, probably associated with the crawling behavior of the larvae (Lakes-Harlan et al. 1991). This finding is in line with the anatomical characteristics of the KO. The organ consists of 3 hair-like and one respectively two papilla-like sensilla. Each sensory dendrite terminates with a tubular body, which strongly indicates mechanosensory function. The arrangement of the sensory hair sensilla indicate that they are deflected by forces from different directions and therefore could be important for orientation. The exploratory behavior of the larva includes straight crawls interspersed with head casts and turns, that the larva executes to redirect the crawling directory. Turns are carried out by backward contractions of the most anterior segments as far as segment four (Berni et al. 2012). Therefore, the strongest directional change appears in the thoracic segments, wherefore the KO could sense these changes in direction. In addition, it would also be imaginable that they are involved in perception of up-and downward movements during orientation behaviors like rearing (Green et al. 1983).

TERMINAL SENSORY CONES

The terminal sensory cones of the last abdominal segments are specialized sense organs, which house mainly hair and knob sensilla. Their prominent cone like structures stick out of the food when the larvae are feeding, putting them in an ideal location to sense changes in oxygen concentrations in their surrounding environment (Vermehren-Schmaedick et al. 2010). Dwelling into the media enables the larvae to hide from predators and strong daylight to prevent desiccation (Kim et al. 2017). But completely submerging into the medium is dangerous as well, as it leads to hypoxia and death, eventually. Therefore, it is essential to dig just deep enough to be sheltered but not completely covered. The cone associated mechanosensory hair sensilla might be activated when the cones are immersing into the medium, and therefore act as a warning signal for the larva.

SPIRACLE SENSE ORGAN

The sense organ of the posterior spiracle consists of four sensilla that sit around the spiracular opening (Figure 17A). They contain one sensory neuron each, with a tubular body at their dendrites’ tip (Figure 17D), which indicates mechanosensory function. The tubular body is surrounded by a dendritic sheath that is connected to electron-dense cuticular material and the spiracular hairs, which sit on top of the spiracle. The hairs protrude in a 90-degree angle from the spiracle (Figure 17 C, D) and are probably deflected when the larva submerges into the substrate. Therefore, the spiracle sense organ could be important for the positioning of the spiracle into the right oxygen-rich environment.

Figure 17:

OUTLOOK

In this work, we further extend the understanding of the larval sensory system. We complete the spatial map of all external sensory structures and internal chordotonal organs. We used ultrastructural details to make well-grounded predictions about the putative sensory functions of these sensory organs. These results will serve as a basis for further molecular and functional studies. To complete the understanding of the different modalities of larval sensation, it will be necessary in the future to also address the somatosensory, pharyngeal, and enteric sensory systems with all their sensilla, cells, and ultrastructure. This can now be done in subsequent studies of the EM volume of the entire body of the larva. A complete description of all external and internal sensory inputs of the Drosophila larva is now within reach.

Material and methods

FLY HUSBANDRY

Drosophila wild type Canton S flies were kept on standard corn meal medium at room temperature like previously described (Selcho et al. 2009; Rohwedder et al. 2012; Apostolopoulou et al. 2014). For all experiments, 3rd instar larvae prior to the wandering stage were used.

SCANNING ELECTRON MICROSCOPY (SEM)

L3 larvae were rinsed with tap water to remove food residuals. Then, specimens were incubated in tap water at 90° C for two minutes. Subsequently, larvae were fixed in 2.5 % glutaraldehyde in 0.1 M HEPES buffer for three hours. The fixative was exchanged each hour. Primary fixation was followed by three washing steps in 0.1 M HEPES buffer, each for ten minutes. Then, samples were incubated in 1 % osmium tetroxide buffered in 0.1 M HEPES for two hours, followed by three washing steps in 0.1 M HEPES for ten minutes each. All steps were performed at pH 7 and 4° C. Next, samples were gradually dehydrated in an increasing ethanol series: 30 %, 50 %, 70 % two times for ten minutes and then 70 % overnight (all at 4° C). Next day, dehydration was continued with 80 %, 90 %, three times 96 % for ten minutes, and 100 % four times for ten minutes (at room temperature). Dehydrated samples were transferred to a critical point drier (Bal-Tec CPD 030, Liechtenstein) and critical point dried over liquid CO₂. Afterwards, samples were glued to aluminum SEM-stubs using adhesive conductive carbon pads (Plano GmbH, Germany) or Conduct-C (Conductive Carbon Cement, Plano GmbH) and subsequently coated with ca. 12 nm platinum in a sputter coater (Quorum Q 150R ES Sputter-Coater, UK). SEM micrographs were acquired on a FESEM Auriga Crossbeam workstation (Zeiss, Germany) using the SmartSEM software-package (Zeiss, Germany).

FOCUSED ION BEAM-SEM (FIB-SEM)

Larvae were rinsed in phosphate-buffered saline (PBS) to remove food residuals. For the study of head organs, the anterior half of the larva was incubated in 2 % formaldehyde fixative (Sigma-Aldrich, Germany) with 2.5 % glutardialdehyde in 0.1 M Na-cacodylate buffer, pH 7.4 for 30-60 min. Then, the head region was cut off and incubated in fresh fixative for another 90 min. For the investigation of thoracic and abdominal segments, the larva was cut into pieces not larger than four segments. Fixation was carried out similar to larval heads. Wash-and post-fixation steps were similar to SEM preparation. En bloc staining was carried out with 1 % uranyl acetate and 1 % phosphotungstic acid in 70 % EtOH in the dark over night before continuing the alcohol dehydration the next day. Samples were transferred to propylene oxide before embedding in Spurr (Plano, GmbH, Germany) using ascending Spurr concentrations diluted in propylene oxide for optimal tissue infiltration (Rist and Thum 2017). Resin was drained of the samples by gravity during polymerization. Polymerization was carried out at 65° C for 72 hours. Samples were mounted on aluminum SEM-stubs using adhesive conductive carbon pads (Plano GmbH, Germany). Samples were sputter-coated with ∼100 nm platinum.

FIB-SEM serial sectioning was carried out using a FESEM Auriga Cross-beam workstation (Zeiss, Germany). Images were acquired using a secondary electron detector or a backscattered electron detector. The three-dimensional FIB-Stack of the DO was acquired using the FIB-SEM operating software package ATLAS 3D Nanotomography (Fibics Inc., Canada).

In FIB-SEM, layers of the sample are mechanically removed by the ion beam of the FIB. This procedure works only for relatively small volumes. Therefore, samples that are embedded conventionally in a block of resin require cutting off excess resin to expose the tissue of interest to FIB-SEM slicing and imaging. Often, this is very difficult, because structures like insect sensilla and sense organs might be too small to be visible through the block surface. Also, the sample lacks contrast as it is dark from the en-bloc-staining with osmium. In this work, we let the samples harden in only minimal amounts of resin instead of embedding the samples in a block of resin (Schieber et al. 2017, Figure S1). The result of this approach were ‘sculptures’ of the samples covered only with a very thin layer of resin that left the surface topology of the larva intact. Immediately, these sculptures could be mounted on EM stubs avoiding further, time-consuming preparation steps. On the minimal embedded sculptures, larger sense organs but also the smallest body wall sensilla could be clearly identified and therefore be precisely targeted by FIB-SEM for serial slicing and imaging.

WHOLE LARVAL VOLUME

Sensilla and neuron reconstruction was done on a STEM (scanning transmission electron microscopy) volume of a whole first instar larva; information on the technical details of its generation are mentioned in Schoofs et al. 2023.

IMAGE PROCESSING

Out of the whole larval volume, we extracted smaller volumes of the sensory organs. The obtained image stacks were imported to Amira (Thermo Fischer Scientific, v2019). Because the whole larval volume was already aligned, the stacks were only slightly realigned by manual correction. FIB-SEM stacks were aligned in FIJI (Schindelin et al. 2012) using the TrakEM2 plugin (Cardona et al. 2012). In Amira, structures of interest were segmented and transformed into 3D objects. Next, the segmentations were imported to Blender (Blender Institute, Amsterdam), where the 3D reconstructions were manually finished using the preliminary segmentation as template.

FIB-SEM/SEM IMAGE PROCESSING

Micrographs were processed in FIJI/ImageJ (NIH, USA) for small contrast and brightness adjustments.

Acknowledgements

This work was supported by the Deutsche Forschungsgemeinschaft (Grant No. 441181781, 426722269, 432195391), by EU funds from the ESF Plus Program (Grant No. 100649752) and by the Open Access Publishing Fund of Leipzig University supported by the German Research Foundation within the program Open Access Publication Funding. We thank Dennis Pauls, Mareike Selcho, Tilman Triphan, Wolf Hütteroth, Denise Weber, Bert Klagges and Reini Stocker for discussions and comments.

Competing interests

The authors declare no competing interests.

Data availibility statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

SUPPLEMENT

Figure S1:

Figure S2:

Figure S3:

Figure S4: