Female and male adults of My. separata were used for the experiments. Larvae were reared on an artificial diet in the laboratory under the conditions of 27°C, 70% relative humidity, and a 16/8 h light/dark cycle. Adults were fed a 10% honey solution.

An environmental scanning electron microscope was used to observe the antennal sensilla. The antennae were fixed in 2.5% glutaraldehyde and then dehydrated with graded ethanol series (50, 70, 90, 96, and 100%). After being mounted on a stub, the antennae were air-dried and gold-coated with sputter-coating before being examined with the electron microscope.

The antennal sensory neurons were mapped by utilizing different staining experiments including (1) dye applied to the base of the scape, visualizing all sensory neurons, (2) dye applied to the base of the pedicel, visualizing all neurons except for those on the scape, (3) dye applied to the base of the flagellum, visualizing the flagellar neurons exclusively, (4) dye applied to cut s. trichoidea on the flagellum, visualizing a selection of olfactory afferents, (5) dye applied to cut sensilla forming the bristle of the pedicel, plus sensilla of the Johnston’s organ, (6) dye applied to sensilla of the bristle on the scape. In all the staining experiments, the adult insect was fixed in a plastic tube with dental wax so that the head was exposed. In the first three experiments, the antenna was cut at the base of the scape, pedicel, and flagellum, respectively, as explained above. Crystals of the fluorescent dye, tetramethylrhodamine dextran (Micro-Ruby, Molecular Probes; Invitrogen, Eugene, OR, United States) were applied at the cut surface by using a needle. In the last three experiments mentioned, dye crystals were applied onto cut sensilla in the relevant regions. After all kinds of staining, the animal was placed in the fridge with a moist filter paper overnight allowing transportation of the dye in the sensory axons. On the second day, the brain and ventral nerve cord was dissected out in Ringer’s saline, and then fixed for 1 h in 4% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS, pH 7.4). Following dehydration with an ascending ethanol series, the preparations were cleared in methylsalicylate, and mounted in Permount in perforated aluminum slides with two glass coverslips. In addition to the staining experiments described above, one experiment of retrograde neural labeling was performed. Here, dye was applied onto the cut end of the antennal nerve close to its entrance into the antennal lobe, in order to visualize the sensory neurons of the Johnston’s organ.

The sensory neurons originating in the LPO were labeled by dye being applied onto the cut tip of the most peripheral segment of one labial palp. The following treatment of the stained preparations was similar to that described in the section above.

For visualizing central neuropil structures in the central nervous system of the moth, some of the mass-stained preparations were labeled by means of synapsin immunocytochemistry as well. After being fixed in 4% paraformaldehyde in PBS and rinsed, as described above, the brain and ventral nerve cord was pre-incubated with 5% NGS (Sigma, St. Louis, MO, United States) in 0.1M PBS containing 0.5% Triton X-100 (PBST; 0.1 M, pH 7.4) overnight at 4°C. The primary antibody SYNORF1 (Developmental Studies Hybridoma Bank, University of Iowa), at a concentration of 1:100 (with 5% NGS in PBST) was then applied, and the preparation was kept at 4°C for 5 days. Following rinse in PBS 6 × 20 min, the brain and ventral nerve cord was incubated in the secondary antibody, Cy2-conjugated anti-mouse (Invitrogen, Eugene, OR; dilution 1:300 with 1% NGS in PBST), for 3 days at 4°C. Finally, the preparation was washed 6 × 20 min in PBS, dehydrated with graded ethanol series (50, 70, 90, 96, each 10 min, and 100% 2 × 10 min), cleared in methylsalicylate, and mounted in Permount, as described above.

A confocal laser scanning microscope (LSM 780, META Zeiss, Jena, Germany) was used to obtain images of the sensory neurons and the neuropils. Two objectives were utilized, a Plan-Neofluar 10 × /0.3 for large-scale images and a 20 × /0.5l for higher resolution. The Cy2, displaying brain structures, was excited by an argon laser of 488-nm line and the Micro-Ruby, displaying the stained neurons, by a HeNe laser 543-nm line. The resolution of the confocal images was obtained at 1024 × 1024 voxels with intervals of 3 or 4 μm.

Identified neuropils within the brain and ventral nerve cord were reconstructed by using the Amira software (Amira 5.3, Visage Imaging, Fürth, Germany). The stained axons and neuronal processes were reconstructed manually as described previously by means of the skeleton tool of Amira (Zhao et al., 2016).

The images from electron microscopy demonstrated that the antennae of My. separata are covered by morphologically different categories of sensilla (Figure 1; Chang et al., 2015). The flagellum is the site for several types of sensilla including s. trichoidea, s. basiconica, s. coeloconica, and s. styloconica. The most abundant category is the s. trichoidea appearing as long hairs (Figures 1A,B) On the pedicel and scape, the Böhm bristles are gathered in clusters at the joints of the pedicel-scape and the scape-head, respectively (Figures 1A,C,E). There are two clusters of Böhm bristles on the pedicel and two on the scape, positioned opposite to each other (Figures 1C,E). The number of bristles on the scape is relatively large. Böhm bristles comprise short s. chaetica (indicated by arrows in Figures 1D,F) surrounded by many short seta (indicated by arrowheads in Figures 1D,F).

Among 12 attempted trials including dye applied to the proximal antennal nerve (AN), for labeling the peripheral part of the antennal sensory neuron, eight were successful. The retrograde staining revealed that the AN is organized into two sub-bundles formed mainly by axons of flagellar sensory neurons (Figures 1G,H). At the proximal part of the pedicel, there is a large number of bipolar neurons confined to the Johnston’s organ (Figures 1G,H,J). The dendrites of these sensory neurons extend into the distal part of pedicel (Figure 1J) whereas their axons join the AN from a surrounding position (Figures 1H,J). The Böhm bristles, which were found both on the scape and pedicel, encapsulate sensory neurons extending their dendrites under the cuticular protrusion (scape: Figures 1G,K; pedicel: Figures 1G,J, indicated by red dotted circle line). Clusters of these sensory neurons are located opposite to each other, two on the scape and two on the pedicel (indicated by arrows in Figures 1H,I). In addition to sensory neurons of Böhm bristles, a few bipolar neurons were observed in the proximal part of the scape (Figures 1G,H,L). These neurons resemble those of the Johnston’s organ by extending their dendrites toward the cuticle and having axons joining the AN (Figure 1H indicated by arrowheads). The AN also includes some motor neurons which innervate intrinsic muscles of the scape (Figures 1M,N).

The gross innervation pattern of antennal sensory neurons in the central nervous system of My. separata was mapped from confocal images of 10 successfully stained preparations (among 21 trials), in which dye was applied to the base of the scape. The axons of all antennal afferents project via the AN to the ipsilateral side of the central nervous system, including the antennal lobe (AL), lateral protocerebrum, AMMC, gnathal ganglion, plus the prothoracic and mesothoracic ganglion of the ventral nerve cord (Figures 2A,C,D and Table 1). Two kinds of additional staining experiments including dye applied to the base of the pedicel (20 trials) and flagellum (24 trials) resulted in 12 and 14 successfully stained preparations, respectively. All these preparations showed labeled projections in the regions mentioned above, including the AL, lateral protocerebrum, AMMC, gnathal ganglion, and ventral nerve cord. However, whereas the specific innervation patterns in the AL and the lateral protocerebrum were similar across preparations stained from the three different antennal segments, they differed in the remaining target regions (Table 1). Males and females displayed identical projection patterns.

The most prominent target area of the antennal afferents was the AL (Figures 2, 3A–C). Here, in the primary olfactory center, the sensory axons form a characteristic pattern consisting of hollow spheres, established by terminals innervating the periphery of the AL glomeruli (Figures 3A–C). All glomeruli receive input from the antennal axons, except for one, which is located most ventrally in the AL (indicated by asterisk in Figure 3B). The projection pattern in the AL was identical in all successfully stained preparations no matter if dye had been applied to the scape, pedicel, or flagellum (Table 1). Specific staining of selected s. trichoidea, situated on the flagellum, showed labeled axons innervating the AL glomeruli exclusively (Figure 3D and Table 1).

In addition to the numerous axons targeting the AL, a few stained fibers projected to the ipsilateral protocerebrum of the brain (Figures 3E,F,H,I and Table 1). This region, which is positioned laterally to the central body and anteriorly of the mushroom body calyx, was innervated by labeled processes projecting via the posterior AL (Figure 3H indicated by an arrowhead). Again, the projection pattern appeared similar in all stained preparations.

In addition to the terminals of sensory neurons, a big soma located at the posterior border of brain, on the ventrolateral side of esophagus, was stained (Figures 3E,G). Its primary neurite bifurcated; one branch targeted the AMMC and the other projected ventrally of the esophagus to the contralateral side (Figure 3I, indicated by an arrowhead). It may be a centrifugal neuron innervating sensilla located on the flagellum. Unfortunately, the terminals of these prominent neural processes could not be traced owing to the weak staining.

A substantial bundle of the antennal axons bypasses the AL on its lateral side and targets the AMMC (Figure 4 and Table 1). According to the pattern of axon terminals, at least eight zones were identified, of which Zones 3, 4, and 8 partly overlap with the gnathal ganglion (Figures 4A–F). Zone 1 is located ventro-posteriorly of the AL, and Zone 2 ventrally to Zone 1 (Figures 4A,D). Zone 3, which is located laterally of the tritocerebrum, close to the dorsal midline of the gnathal ganglion, receives projections of antennal axons bypassing Zone 1 (Figures 4A,D). Zone 4 is located medio-posteriorly to the ventral part of Zone 2 and receives projections of axons bypassing Zone 1 (Figures 4B,D). Zone 5 is located posteriorly of Zone 2 (Figures 4B,E), Zone 6 medio-posteriorly of the dorsal AMMC, and Zone 7 medio-posteriorly of the ventral AMMC (Figures 4C,F). Zone 8, which borders with the midline of the gnathal ganglion and receives projections bypassing Zone 7, is located ventro-medially of Zones 4 and 7 (Figures 4C,D).

The preparations stained at the base of the scape and the pedicel showed similar projections patterns including labeled axons terminating in all eight zones of the AMMC (Figures 4A1–F1,A2–F2). The preparations stained at the base of the flagellum, on the other hand, showed labeled axons terminating in only six zones, i.e., Zones 1 – 4, plus 7 and 8 (Figures 4A3-F3).

Specific staining of Böhm bristles demonstrated that their sensory axons innervate almost the same areas, i.e., Zones 4, 5, 7, and 8, whereas specific staining of sensilla on the Johnstons’ organ gave rise to labeled terminals in Zones 4 – 8 (Figures 4G–K and Table 1). The preparations stained from the base of the scape showed a few labeled somata as well (Figure 4A1). According to previous reports from Ma. sexta, these are probably motor neurons (Kloppenburg et al., 1997). As demonstrated in Figures 4D1–F1, these neurons arborize mainly in a region located ventro-posteriorly in the AMMC (Zone 9). Notably, none of these motor neurons were stained when dye was applied at the base of the pedicel and the flagellum.

Within the axon bundle bypassing the AL, there are some long fibers forming two descending tracts that project into the ventral nerve cord. These paths, named T1 and T2, are separated when they bypass Zone 4 and 8, but fuse before entering the ventral nerve cord (Figures 4D–F).

In the ventral nerve cord, the antennal axons project via the ipsilateral connective, first to the prothoracic ganglion (proTG) and then to the mesothoracic ganglion (mesoTG, Figure 5 and Table 1). In the proTG, the axons pass near the midline and give off processes in the central area of the neuropil (Figure 5A). A few processes cross the midline to the contralateral side and some project to the lateral part of the neuropil (Figure 5B, indicated by arrows). The axons project further to the central area of the mesoTG (Figure 5C). Here, the axons make a loop and at the same time give off processes. As in the proTG, a few processes cross the midline to the contralateral side and some project to the lateral part of the neuropil (Figure 5D, indicated by arrows).

The preparations stained at the base of the scape and the pedicel showed similar projection patterns in the ventral nerve cord (Figures 5A1–D1,A2–D2), whereas those stained at the base of flagellum showed labeling of some axons of the T1 tract only. This fiber bundle, which is substantially thinner than the whole T1 tract, can be observed as it exits the gnathal ganglion and projects to the ventral nerve cord (Figures 4D3–F3, 5A3–D3).

Among 33 preparations attempted labeled by applying dye into the LPO, 16 were successful. The stained axons of the LPO sensory neurons projected via the ipsilateral labial palp nerve and terminated in three main areas of the central nervous system: (1) LPOG in each AL, (2) the gnathal ganglion, and (3) the proTG of the ventral nerve cord (Figure 6). The most prominent target region was the LPOG. After entering the gnathal ganglion, a substantial portion of the stained axons divided into two bundles, the antenno-gnathal tracts, each projecting to the LPOG in one AL (Figures 6A,E,F). Here, the stained terminals filled the whole LPOG, and a few processes even extended outside the glomerulus (Figures 6C,D, indicated by arrows). Staining of antennal sensory neurons demonstrated that the LPOG does not receive any input from this neuronal category (Figure 3B, indicated by an asterisk). The second target region of LPO axons, the gnathal ganglion, is innervated by processes terminating mainly in the ipsilateral neuropil. Only a few stained branches extended to the contralateral side (Figures 6B,E, arrowheads). The third target region, the proTG of the ventral nerve cord, was innervated by axons leaving the gnathal ganglion via the ipsilateral connective (Figures 6G–I). In the proTG, the stained axons terminated in the anterior part of the neuropil forming a loop (Figure 6G). Males and females showed identical staining patterns in all three regions of the central nervous system.

Data obtained by electron microscopy showed that a large number of sensilla, including different morphological categories, are present on the flagellum of My. separata, whereas the pedicel and scape carry mainly Böhm bristles. These data are in full agreement with previous reports on the same species (Chang et al., 2015) and other insects (Schneider, 1964; Keil, 1999). The retrograde staining experiments visualized the antennal sensory neurons housed by the sensilla and their axons forming the AN. In addition, a large number of JONs located close to the inner surface of the pedicel, plus a few bipolar neurons at the inner surface of the scape were observed. Since no obvious sensilla were found on the external surface of the scape, these bipolar neurons may act as proprioceptors for the scape. Their specific function remains to be explored, however. Interestingly, the Böhm bristles and JONs are similar to those of Ma. sexta, which are found to work as mechanosensory neurons detecting movement and position of the antenna (Sane et al., 2007; Krishnan et al., 2012; Dieudonné et al., 2014). Thus, they are probably involved in flight control (Sane et al., 2007). The antennae of the monarch butterfly possess magnetic sensors and a circadian clock facilitating optimal orientation during migration (Merlin et al., 2009; Guerra et al., 2014). My. separata might actually possess corresponding arrangements as well, since it is reported to maintain consistent directions when migrating (Xu et al., 2017). Generally, the antennal sensilla of My. separata comprise several morphological types designed to detect different modalities, which in turn allow My. separata to perceive simultaneous input about the complex environment.

By combining electron microscopy with different mass staining experiments, including dye applied to the base of the scape, pedicel, and flagellum, respectively, we obtained a general overview of all antennal projections in the central nervous system as well as more detailed maps visualizing the target regions of sensory axons housed inside sensillum types located on the different antennal segments. The general staining pattern of all antennal afferents covered innervations in several ipsilateral regions of the central nervous system including the AL, protocerebrum, AMMC, gnathal ganglion, proTG, and mesoTG. Such a widespread projection pattern of the antennal axons was also observed in the locust, Locusta migratoria, blowfly, Calliphora erythrocephala, blood-sucking bug, Rhodnius prolixus, and mirid bug, Apolygus lucorum (Bräunig et al., 1983; Nässel et al., 1984; Barrozo et al., 2009; Xie et al., 2016). The multiple targeting areas may correspond to the different sensory modalities being detected by the antennal sensory neurons. Such a multisensory organ as the antennae of My. separata provides the flying insect with detailed information about the external world.

In addition to the afferent axons, a putative centrifugal neuron innervating the antenna was stained as well. This kind of neuron has not been reported in any insect species so far.

The finding of stained LPO axons entering the ipsilateral side of the gnathal ganglion via the labial nerve and projecting to three distinct areas, the LPOG, gnathal ganglion, and proTG, is similar to the projection pattern previously reported in Helicoverpa armigera and other species (Bogner et al., 1986; Kent et al., 1986; Zhao et al., 2013). Several studies on moths have demonstrated that the sensory neurons in the LPO respond to CO₂ (Bogner, 1990; Stange, 1992; Stange et al., 1995; Guerenstein et al., 2004). It is therefore reasonable to assume that the LPO neurons of My. separata are involved in CO₂ detection as well. In particular, it would be interesting to investigate whether this kind of signaling might be involved in the long-distance migration of My. separata, taking place at altitudes of 200 – 500 m (Zhang et al., 2013).

The staining pattern in the LPOG of My. separata, including terminals from LPO axons filling the whole glomerulus, and no antennal axons, is in full agreement with previous reports in other moth species (Kent et al., 1986; Bogner, 1990; Zhao et al., 2013). Correspondingly, in the mosquitos, Ae. aegypti, and An. gambiae, CO₂ sensitive neurons located on the distal segment of the maxillary palps project to AL glomeruli that receive no input from antennal sensory neurons (Distler and Boeckh, 1997; Anton et al., 2003). The finding of a few processes extending outside of LPOG in My. separata is not reported in any of the other studied species.

As mentioned above, the target regions of LPO projections in the gnathal ganglion and proTG of My. separata, were also observed in Helicoverpa armigera (Zhao et al., 2013). Also, previous findings from these two moth species have reported two morphologically corresponding types of LPO sensilla, one hair-shaped and one club-shaped (Zhao et al., 2013; Dong et al., 2014). In future studies, it would be interesting to map the projection pattern of axons originating from each sensillum category.