Male H. armigera pupae (Lepidoptera; Noctuidae, Heliothinae), purchased from China (Henan Jiyuan Baiyun Industry Co., Ltd), were kept in climate chambers (Refritherm 200 and 6E, Struers-Kebo lab, Albertsund, Denmark, or Binder KBF 720, Tuttlingen, Germany) at 24°C and 70% air humidity on a reversed night-day cycle (14 h light and 10 h dark). The moths were supplied 10% sucrose solution. Experiments were conducted 1–3 days after the emergence. According to Norwegian law of animal welfare there are no restrictions regarding experimental use of Lepidoptera.

In all experiments, the insect was placed inside a 1 ml plastic pipette (VWR chemicals, France) with the head exposed and then immobilized with dental wax (Kerr Corporation, Romulus, MI, United States). The outer scales of the third labial-palp segment were carefully removed. In order to determine the LPO specific projections, three types of mass staining experiment were performed. To label the projections of the LPO sensory neurons exclusively, the outer wall of the third segment of the labial palp was first sealed with Vaseline. After that, we applied a localized injection by inserting a blunt glass electrode filled with 4% Micro-Ruby (biotinylated dextran-conjugated tetramethyl rhodamine, Molecular Probes) in 0.2 M potassium acetate (KAc) solution into the LPO, with depolarizing current pulses of 7–8 nA at 1 Hz for 2 min. The presence of non-LPO sensory neurons on the labial palp was determined by applying crystals of fluorescent dye on two distinct sites on the outer surface of the third labial-palp segment: (i) a small longitudinal cutting site carefully made on the cuticle of the outer wall and (ii) a transverse cutting site at the very tip of the labial palp.

For investigating the projection pattern of sensory cells housed in the hair-shaped (distal part) and the club-shaped (basal part) sensilla, two kinds of anterograde mass-staining experiments were conducted by cutting at different sites of the labial-palp segment. We first obtained an overview of all the sensory neurons in both sensillar types of the LPO by applying crystals of Micro-Ruby to the transverse cutting plane at its basal part. To label the sensory cells housed in the hair-shaped sensilla exclusively, crystals of the fluorescent dye were applied at the transverse cutting plane made on the intermediate part of the LPO. After staining, the insects were kept for three days in dark to allow anterograde axonal transportation of the dye. A dissecting microscope (Leica M60) equipped with a CCD camera (Leica DMC 4500) was used to determine the location of the transverse cuts made for dye application.

In addition, double-labeling experiments including application of two fluorescent dyes were carried out. One dye was allocated to the intermediate part of the LPO and one to the basal. First, the LPO was sectioned at the intermediate part and then exposed to crystals of Alexa 488. After 24 h at 4^oC in a humid chamber, the LPO was sectioned at the base and exposed to Micro-Ruby. The insects were then kept for two more days in dark before the brains were dissected.

All brains were dissected out in Ringer’s solution [in mM: 150 NaCl, 3 CaCl₂, 3 KCl, 25 Sucrose, and 10 N-tris (hydroxymethyl)-methyl-2-amino-ethane sulfonic acid, pH 6.9] before being fixed in a paraformaldehyde solution (4% PFA in 0.1 M phosphate buffer, pH 6.9) for 2 h at room temperature or overnight at 4^oC. After fixation, the brains were dehydrated in an ascending ethanol series (50, 70, 90, 96, and 2 × 100%; 10 min each), and finally cleared in methyl salicylate (Sigma-Aldrich, Germany). Preparations were then mounted in permount (Chemi-Teknik As, Oslo).

In order to investigate the projection pattern of individual LPO sensory neurons, we first performed iontophoretic staining experiments from the LPO. The insect was immobilized in a plastic pipette as described above. A sharp quartz electrode pulled by a laser-based horizontal puller (P – 2000; Sutter instruments, CA, United States) was used for dye injection. This capillary which contained a fluorescent dye solution (4% Micro-Ruby in 0.2 M potassium acetate solution) was backfilled with 0.2 M KAc and inserted into the LPO via a micromanipulator (Leica microsystems, Wetzlar, Germany). The recording electrode had a resistance of 100–200 MΩ. In addition to anterograde labeling, we performed retrograde labeling by accessing individual sensory neurons projecting to the LPOG. The head cuticle between the eyes was then removed by using a sharp razor-blade knife. The brain was exposed by taking away muscles and tracheas. Likewise, a quartz electrode with dye was inserted into one of the LPOGs. In both types of iontophoretic staining experiments, a chloridized silver wire placed in the eye served as a reference electrode. When the neuron contact was stable, injection of the fluorescent dye was induced via 200 ms pulse depolarizing currents of 8–10 nA for 15 min. After staining, the preparation was kept one to three days at 4^oC and then dissected, fixed, dehydrated, cleared, and mounted in methyl salicylate.

All successfully stained neurons were imaged by using a confocal laser-scanning microscope (LSM 800 Airyscan, LSM 800 GaAsp -Pmt-1, Zeiss, Jena, Germany) equipped with a C- Apochromat 10×/0.45 water objective, Plan-Neofluar 20x/0.5 air objective, and LD LC1 Plan – Apochromat 25×/0.8 imm Korr DIC M 27. The Micro-Ruby (Ex_max 555 nm/Em_max 580 nm) labeling was excited by the 543-nm line of a HeNe laser, and Alexa 488 (Ex_max 495 nm/Em_max 519 nm) staining by the 488-nm line of an argon laser. For the samples stained with Micro-Ruby, we used two channels: one for exciting the Micro-Ruby and the other, the 488-nm line of an argon laser, for detecting brain neuropils, made visible via autofluorescence from the sample. We used two detectors: airyscan detecting Micro-Ruby and GaAsP for detecting the autofluorescence. These detectors were used for the double-labeled preparations as well. High-resolution confocal images with 1,024 × 1,024 pixels, at distances of 2 to 6 μm in the z-direction were obtained. The pinhole size was 1 airy unit and the pinhole diameter 36 μm. Images were collected and saved as 8-bit. czi files. In order to quantify the fluorescence intensities in the two LPOGs during dye application to the intermediate and basal segment of the LPO, the 10X water immersion objective with 0.45 NA was used. The image pixel size was set to 0.209 μm and the zoom to 0.8. The images were scanned at 0.55 μs pixel dwelling time. The detector gain was set to 530–630 V, the digital offset to 0, and the digital gain to 1. The axial distance (z thickness) was chosen according to the optimal slicing, i.e., 4.98 μm (having at least 50% overlap of two successive optical sections).

All the confocal images were processed in the Zen 2.3 software (Blue edition, Carl Zeiss Microscopy GmbH, Jena, Germany). To estimate the density of sensory projections in the ipsilateral and contralateral LPOG, the fluorescence intensity of each LPOG was quantified by using Image J¹. Furthermore, in order to compare the staining patterns formed by the two types of sensory neurons housed by hair-shaped and club-shaped sensilla, eight preparations labeled from the base of the pit and six from the intermediate part were included in the analysis. First, we selected the relevant part of the confocal stack, i.e., the images containing the LPOGs. The fluorescence intensities of the LPOGs were then registered through measuring the fluorescence in the glomerular region within the stack. The background fluorescence, i.e., the autofluorescence, was quantified in a corresponding manner from the region between the two LPOGs of the same stack. The LPOG fluorescence intensity per square micrometer was then normalized by subtracting the background fluorescence: mean fluorescence intensity within the LPOG area (μm²) minus mean fluorescence intensity within the randomly selected background area (μm²). Since the data were not normally distributed, the non-parametric Wilcoxon signed-rank test was performed to analyze the data within subjects. All probabilities given were two-tailed. For analyzes of data linked to the two cutting levels, obtained from different individuals, the Mann-Whitney U test was utilized. Effect size (r) was calculated by dividing the standardized test statistics by the square root of the sample size (Pallant, 2007). The Statistical package for the social sciences (SPSS), version 25, was used for statistical analysis.

Iontophoretic staining experiments performed by inserting a blunt glass electrode into the pit showed that neurons originating from sensilla located inside the LPO project exclusively to the antennal lobes. Here they target the ventrally located LPOG. As shown in one of totally five successfully labeled preparations, an axon bundle projecting via the labial nerve can be seen (Figure 1A). After entering the ipsilateral part of the GNG, it passed on dorsally, close to the midline ipsilaterally in the GNG. About 70 μm ventrally of the esophagus, it divided into two branches, each targeting the LPOG in one antennal lobe. In addition to this main axon bundle projecting ipsilaterally in the GNG, another thin sub-bundle crossed the midline about 170 μm ventrally of the esophagus and projected dorsally on the contralateral side (Figure 1A). Confocal images of preparations stained from both sides, showed that this thin sub-branch merged with the main fiber bundle from the other LPO (Supplementary Figure S1). None of the fiber bundles extended terminal branches in the GNG. The four remaining preparations successfully stained by the same technique showed a similar projection pattern including a denser innervation in the ipsilateral LPOG than in the contralateral one (Figure 2B and Supplementary Figure S2).

To determine the projection patterns of sensory neurons located outside the LPO, we first conducted focused mass staining by applying dye to a longitudinal section carefully made on the outer wall of the third labial-palp segment (Figures 1B,E). Five successfully labeled preparations showed stained fibers entering the GNG via the labial nerve, but unlike the projection pattern of sensory neurons originating inside the pit, no projections targeted the LPOGs. As shown in Figure 1B, the main portion of projections from the outer wall terminated in the ipsilateral GNG. In addition to these innervations, a bundle of fibers projected to the ventral nerve cord and a few axons targeted the ipsilateral antennal mechanosensory and motor center (AMMC). The four remaining preparations stained in the same manner showed similar projection patterns (Supplementary Figure S3). In one preparation, a very weak staining can be observed in the LPOGs – possibly due to dye leakage from the outer cuticle (Supplementary Figure S3E). An additional series of mass staining experiments from the palp surface were carried out by applying dye onto the distal cross section of the third segment (n = 5; Figure 1C). Again, a bundle of stained axons confined to the labial palp nerve innervated the GNG. However, unlike the targets of axons originating from the longitudinal section of the palp surface, these fibers terminated not only in the ipsilateral GNG but also the contralateral. Besides, there were only a few axons projecting to the ventral nerve cord and no innervation of the AMMC (Figures 1C,E and Supplementary Figure S4). Noticeably, combining the three projection patterns obtained by applying dye into the LPO and to the two different regions of the palp surface, gives a total pattern very similar to that obtained when applying dye to the transverse section of the peripheral labial-palp, at the proximal level (Figure 1D).

We performed iontophoretic staining with a sharp glass electrode both from the LPO and the LPOG, providing morphological identification of individual LPO neurons via anterograde and retrograde labeling, respectively. The assembly of successfully stained neurons comprised three morphological types – all entering the GNG via the labial nerve and then projecting to the LPOG in the antennal lobe. The first type comprised bilateral neurons projecting to the LPOG in both antennal lobes (Figures 2A–E). The second type was unilateral, targeting the LPOG in the ipsilateral antennal lobe exclusively (Figure 2B). The axons of these two neuron types followed the main fiber bundle ipsilaterally in the GNG. The third neuron type had an axon crossing the midline of the GNG and targeting the LPOG in the contralateral antennal lobe only (Figures 2C,D). This contralateral neuron type included two sub-types. One sub-type followed the main axon bundle on the ipsilateral side of the GNG before crossing the midline and terminating in the contralateral LPOG (Figure 2C) whereas the other sub-type targeted this glomerulus via the thin bundle on the contralateral side of the GNG (Figure 2D, n = 2). No other areas of the central nervous system were innervated by the individual neurons.

To investigate whether the two morphological types of LPO sensilla (hair-shaped sensilla located distally and club-shaped located basally) house sensory neurons having different projection patterns, mass staining was made from two different sites of the peripheral palp. Fluorescent dye was applied to the transverse cut made at the intermedial and basal part of the third segment, respectively (Figure 3). This classic mass staining labeled both the LPO-specific sensory neurons housed inside the pit and non-LPO sensory neurons located on the outer cuticle and tip of the labial palp. Tracing of the LPO sensory neurons from the basal and intermediate part of the palp demonstrated similar projection patterns in the form of stronger fluorescence intensity in the ipsi- than the contra-lateral LPOG (Figures 3A–C). In the preparations stained from the base (n = 8), including both sensillum categories, the mean fluorescence strength was significantly higher within the ipsilateral LPOG (Mdn = 65.071) than in the contralateral (Mdn = 24.76, z = 2.52, p = 0.012, r = 0.89; Figure 3D). (ii) Likewise, the samples stained at the intermediate part (n = 6), including hair-shaped sensilla only, showed a significantly stronger fluorescence intensity in the ipsilateral LPOG (Mdn = 25.34) than in the contralateral (Mdn = 8.27, z = 2.20, p = 0.03, r = 0.89; Figure 3D). In addition, the mean intensity in the ipsilateral LPOG in the preparations stained from the base was significantly different from the corresponding fluorescence intensity in the preparations stained from the intermediate level (Mann-Whitney U test, U = 48, z = 3.098, p < 0.01, r = 0.59; Figure 3D). Likewise, the mean intensity in contralateral LPOG in the preparations stained from the base and the intermediate part of the LPO was significantly different (U = 41, z = 2.91, p < 0.05, r = 0.82; Figure 3D).

Two of the preparations that were stained from the base of the third segment of the labial palp demonstrated antennal-lobe projections having a few branches ramifying outside the LPOG. One preparation showed a short axon projecting to an ordinary glomerulus located postero-medially of the LPOG in the ipsilateral antennal lobe (square box in Figure 4A). The other preparation included a sensory neuron forming an axonal loop dorsally of the LPOG in the ipsilateral antennal lobe, but without targeting any other glomerulus (Figure 4B).

Selective staining of sensilla housed inside the pit, carried out for the first time, demonstrated that the LPO sensory neurons of H. armigera target the antennal lobes exclusively. This refines the previous findings on central projections of LPO neurons in the current species, reporting about two target areas in addition to the antennal lobes, i.e., the GNG and the ventral nerve cord (Zhao et al., 2013). In the former study, however, a more unspecific staining technique including dye application to the transverse cut surface of the peripheral palp was utilized. Distinct staining of sensilla located on the outer palp surface, as performed here, revealed that the projections originating from this site terminate in the two additional areas – the GNG and the ventral nerve cord. “Merging” the staining patterns obtained from the two labeling techniques utilized in the present study (Figures 1B,C), forms a pattern that is very similar to that previously reported by several researchers using the classic mass staining technique (Figure 1D; Kent et al., 1986; Lee and Altner, 1986; Zhao et al., 2013; Ma et al., 2017). This confirms the suggestion made by Kent et al. (1986) that the axon projections to the LPOGs originated from sensory neurons housed inside the pit and that the branches seen in the GNG were parts of other sensory neuron types. Interestingly, a short peg sensillum type responding to humidity was recently identified on the labial palp of the Asian longhorned beetle, Anoplophora glabripennis (Hall et al., 2019).

The fact that the LPO sensory neurons target the LPOG in each antennal lobe, and no other areas of the central nervous system, means that the projection pattern of CO₂ sensitive neurons in the moth is quite similar to that found in other insects, including flies and mosquitoes. For example, in the fruit fly, D. melanogaster, the CO₂ sensitive neurons project to a ventrally located antennal-lobe glomerulus, called the V glomerulus (Stocker et al., 1983; Suh et al., 2004), and in the mosquito, Aedes aegypti, the corresponding neuron type targets one dorsomedial glomerulus in the antennal lobe (Distler and Boeckh, 1997; Anton et al., 2003; Anton and Rospars, 2004). These two species have ipsilateral projections innervating only one antennal lobe, whereas the moth has a substantial portion of bilateral projections targeting the LPOG in both antennal lobes. Remarkably, yet another well-known mosquito species, Anopheles gambiae, has bilateral innervation – like the moth (Anton et al., 2003; Anton and Rospars, 2004).

Notably, there seem to be obvious differences in the peripheral arrangement of the CO₂ pathway across the various species. Moths have sensilla dedicated to CO₂ sensory neurons alone (Zhao et al., 2013). Fruit flies and mosquitoes, on the other hand, have their CO₂ sensitive neurons co-located with olfactory sensory neurons inside olfactory sensilla on the antenna and the maxillary palp, respectively (Riesgo-Escovar et al., 1997; Grant and O’Connell, 2007; Lu et al., 2007).

Previous and present mass staining experiments on lepidoptera, including M. sexta, Pieris rapae, and H. armigera, have demonstrated that both LPOGs receive input from sensory neurons located within each LPO (Kent et al., 1986; Lee and Altner, 1986). Since only a few individual sensory neurons were previously stained (Guerenstein et al., 2004), knowledge about these neurons’ morphologies has been lacking. The three sub-bundles which originate from the labial nerve and project to the LPOGs, as shown in Figures 1A,D, already indicate the three morphological types of LPO sensory neurons identified here: one bilateral, one unilateral, and one contralateral type. To our knowledge, no contralateral CO₂ sensory neuron has been reported in any other insect order. Since contralateral projections from olfactory or gustatory sensory neurons have never been reported in any insect species, the LPO projections targeting the contralateral LPOG in the moth brain seem to be unique.

Considering the large number LPO sensory neurons, which are housed inside each pit (ca. 1200 sensilla) in H. armigera (Zhao et al., 2013), the signal-to-noise ratio in the single target glomerulus is particularly high. The fact that there is a substantial proportion of bilateral projections makes this value even higher. Such an input arrangement might be optimally designed for detecting minor CO₂ fluctuations (Stange, 1992; Guerenstein et al., 2004). Previous reports have shown that the honey bee and the fruit fly perform simultaneous comparison of the odor concentrations detected by the two antennae in order to trace the source (osmotropotaxis; Martin, 1965; Borst and Heisenberg, 1982; Stocker et al., 1990). Due to the very strong convergence in the CO₂ system of the moth, we assume a similar mechanism even though the two palps are very closely located.