Neuronal architecture of the second-order CO2 pathway in the brain of a noctuid moth

Many insects possess the ability to detect fine fluctuations in the environmental CO2 concentration. In herbivorous species, plant-emitted CO2, in combination with other sensory cues, affect many behaviors including foraging and oviposition. In contrast to the comprehensive knowledge obtained on the insect olfactory pathway in recent years, we still know little about the central CO2 system. By utilizing intracellular labeling and mass staining, we report the neuroanatomy of projection neurons connected with the CO2 sensitive antennal-lobe glomerulus, the labial pit organ glomerulus (LPOG), in the noctuid moth, Helicoverpa armigera. We identified 15 individual LPOG projection neurons passing along different tracts. Most of these uniglomerular neurons terminated in the lateral horn, a previously well-described target area of plant-odor projection neurons originating from the numerous ordinary antennal-lobe glomeruli. The other higher-order processing area for odor information, the calyces, on the other hand, was weakly innervated by the LPOG neurons. The overlapping LPOG terminals in the lateral horn, which is considered important for innate behavior in insects, suggests the biological importance of integrating the CO2 input with plant odor information while the weak innervation of the calyces indicates the insignificance of this ubiquitous cue for learning mechanisms.

www.nature.com/scientificreports/ to determine the main output regions of the LPOG neurons identified here (Fig. 2a, b). All newly reconstructed regions are listed and presented in Fig. 2c.

The PN pathway carrying CO 2 input is distinct from the canonical olfactory PN pathway.
In order to obtain an anatomical overview of the PN pathway carrying CO 2 information in the moth brain, a series of focused mass staining experiments from the LPOG was first performed. Confirmation about dye application into the LPOG, and the LPOG exclusively, was obtained by examining the presence of labeled sensory neurons in the LPO (see confocal scanning in Supplementary Fig. S1) and a lack of staining in the ventral AL glomeruli, located adjacent to the lateral cell body cluster (LC). As demonstrated in one of the preparations, including the highest number of simultaneously labeled LPOG PNs (Fig. 3a), the projection pattern demonstrated extensive labeling in the LH. This pattern originated from overlapping terminals of axons passing along three distinct ALTs, the transverse, medial, and lateral tract. The transverse-tract neurons were particularly prominent. The relatively uncommon transverse-tract neuron type is previously reported to carry information from a few AL glomeruli only 26 , and our data show that the LPOG is one of them. In addition to targeting the LH, some LPOG output neurons also projected to the Ca, however, with substantially weaker innervations. In order to compare the projection patterns of the LPOG PNs and a typical medial-tract PN, we placed confocal images of the two PN categories next to each other (Fig. 3a, b). The considerably stronger staining of the Ca in the medial-tract PN is obvious. Furthermore, the mass stained preparation in Fig. 1, including AL output neurons originating www.nature.com/scientificreports/ from seemingly all glomeruli, displays extensive innervations in the Ca as well. To quantify the characteristic difference in the terminal patterns of LPOG PNs versus PNs originating from ordinary glomeruli, we measured the fluorescence intensity within the Ca and the LH region in each of the three stained preparations referred to above. The Ca/LH-ratios were then calculated and found to be 0.42 for the LPOG PNs (Fig. 3a), 1.35 for the AL mass-stained preparation (Fig. 1), and 0.89 for the single medial-tract PN (Fig. 3b). Thus, we assumed that the LPOG PNs were relatively weakly connected to the Ca. Next, to confirm the contrasting distribution of axon terminals in the LH vs. Ca of LPOG PNs, we carried out two types of retrograde mass staining experiments. We first injected fluorescent dye into the Ca, which resulted in extensive staining of all glomeruli in the ipsilateral AL except for one, the LPOG (Fig. 3c). By quantifying the mean fluorescent intensity in the ipsilateral LPOG and the neighboring glomerulus, G71 ( Fig. 3c and Supplementary Fig. S2), we found that the LPOG was consistently weaker stained (t(5) = 4.97, p = 0.004). This confirmed our hypothesis of a relatively weak connection between the LPOG and the Ca. The second retrograde mass staining experiment involved application of dye into the lateral protocerebrum, including LH. This caused not only labeling of all glomeruli in the ipsilateral AL, including the LPOG, but also uncovered an assembly of stained axons linked to the LPOG in the contralateral AL (Fig. 3d). Here, nine labeled LC somata connected with the innervated glomerulus could be seen-indicating the presence of at least nine bilateral or contralateral LPOG PNs. Altogether, the results from the three kinds of mass staining experiments revealed that the LPOG, receiving input from the labial palp, has an output pathway differing substantially from that of PNs linked to the other AL glomeruli, receiving input from the antennae. The main characteristics of the LPOG pathway include: 1) a prominent involvement of the transverse ALT, 2) relatively weak terminal innervations of the Ca, and 3) a significant proportion of bilateral/contralateral PNs. www.nature.com/scientificreports/ LPOG neurons are uniglomerular PNs with rich morphological diversity. As the two PN pathways, devoted to CO 2 and ordinary odorants, respectively, display substantially different projection patterns at the neuron population level, we next aimed at investigating how these differences were expressed at the singleneuron level. We thus carried out intracellular dye injection into the thick dendrites of LPOG PNs. Fifteen LPOG neurons were morphologically identified, all having glomerular dendritic arborizations focused in the LPOG solely. These neurons were confined to five different tracts: ten to the transverse ALT, two to the medial ALT, one to the lateral ALT, one to the dorsomedial ALT, and one to the antenno-suboesophageal tract (AST). Naming of neuron types and subtypes is an adaptation of the system used in previous studies 18,26,27 . This implies that the five neuron types were named according to the tract they projected along (i.e., Pt, Pm, Pl, Pdm, and Past) whereas the neuron subtypes were classified according to their axonal projection patterns in the protocerebrum. For the name, Pt_a (LPOG) , "P" indicates the PN category, "t" applies to the transverse-tract type, "a" refers to the PN subtype, while " (LPOG) " represents the uniglomerular dendritic arborization. An overview of the individually stained neurons and their corresponding output regions is shown in Table 1 and Fig. 4.
The main share of stained LPOG neurons passed along the transverse antennal-lobe tract. Even though relatively few projection neurons are reported to connect the AL to the protocerebrum via the transverse tract, the main proportion (more than 65%) of labeled neurons originating from the LPOG passed along this route. The transverse-tract neurons were morphologically diverse, including both uni-and bilateral neurons. They all left the AL together with the medial ALT and projected along its path until the posterior edge of the CB. Here, they bent off from the medial tract, passing along the pedunculus before projecting their wide-spread terminal branches in different areas of the ipsilateral and/or contralateral protocerebrum. The transverse-tract neurons comprised four subtypes, two unilateral and two bilateral (Fig. 5).
Unilateral transverse-tract neurons. Four of the ten labeled transverse-tract LPOG PNs were unilateral. Three of these unilateral neurons were morphologically similar forming one distinct subtype and the fourth constituted another subtype. The three first-mentioned neurons were the only ones innervating the Ca. In addition, these neurons projected to several regions in the lateral protocerebrum, including the LH, superior lateral protocerebrum (SLP), posterior lateral protocerebrum (PLP), and ventrolateral protocerebrum (VLP, Figs. 4d, 5a). This neuron subtype which was named Pt_a (LPOG) , is similar to an LPOG PN previously described in M. sexta 34  LPOG neurons confined to the medial antennal-lobe tract bypass the calyces (Ca). One of the notable features of the mass-stained LPOG PNs was the weak innervation of the Ca-even when multi- www.nature.com/scientificreports/ ple medial-tract neurons were labeled (Fig. 3a). The single-neuron staining experiments affirmed this feature. The two medial-tract LPOG neurons individually stained here, formed two subtypes-both bypassing the Ca (Figs. 4h, i and 6a, b). One unilateral medial-tract neuron originating in the LPOG bypassed the Ca but innervated various areas of the protocerebrum including the LH, SLP, PLP, and VLP (Figs. 4h, 6a). In addition, the axon extended a few side-branches into the inferior neuropil (INP). We classified this subtype as Pm_e (LPOG) , in accordance with previously reported neurons displaying similar morphologies 27 . The other medial-tract neuron originating in the LPOG projected to the contralateral protocerebrum (Figs. 4i, 6b). The axon of this subtype passed along the initial part of the ipsilateral medial ALT, however, then it turned and passed along the posterior edge of the CB, crossing the brain midline and targeting the contralateral LH exclusively. This new subtype was named Pm_g (LPOG) .
A lateral-tract neuron connecting the LPOG directly to the LH. One of the 15 labeled LPOG output neurons projected along the lateral ALT. This neuron targeted various regions in the ipsilateral protocer- (Spike activities during stimulation with fresh air and CO 2 is shown to exemplify how LPOG neurons were identified during the experiment.) (b1, b2) 3D reconstruction and confocal image of a Pt_d (LPOG) neuron subtype. The ipsilateral neuron originated in the LPOG and projected to the superior medial protocerebrum (SMP) and SLP without innervating the Ca. (c1, c2) 3D reconstruction and confocal image of a Pt_e (LPOG) neuron subtype. This bilateral neuron arborized in the LPOG of both ALs and terminated in the LH contralateral to the site of the cell body. The two LPOGs were connected via the antennal-lobe commissure. The soma was in the medial cell body cluster. The neuron was co-stained with a weakly labeled lateral-tract LPOG neuron. The two insets in c2 are confocal images showing the dendritic arborizations in each AL. The LPOG is located adjacent to glomerulus G71. (d1, d2) 3D reconstruction and confocal image of a Pt_f (LPOG) neuron subtype. This bilateral LPOG neuron terminated in the ipsilateral LH and in the VLP of both hemispheres. The neuron was co-stained with a Pt_d (LPOG) neuron and a lateral-tract LPOG neuron. (The reconstruction of neuronal dendrites in b1, c1, and d1 are shown for illustrative purposes only. Due to prestaining/mass staining of the LPOG, the detailed arborizations were not visible.) All images are in dorsal orientation. CB, central body; A, anterior; L, lateral; P, posterior; Scale bars: 50 μm. LPOG neurons confined to minor tracts. In addition to the three LPOG neuron types described above, including PNs confined to the transverse, medial, and lateral ALT, the single neuron labeling technique explored two rare types, which are not frequently stained. One of these LPOG neurons projected in the dorsomedial ALT (dmALT), which is a relatively thin fiber bundle passing more dorsally and medially than the prominent medial tract. Whereas previously described neurons confined to the dorsomedial tract are reported to project bilaterally 26 , the neuron identified here projects unilaterally. From the LPOG, the main axon passed along the dmALT to the SMP, INP, and the ventro-medial neuropil (VMNP, Figs. 4k, 6d). In addition, a few fibers projected to the antennal mechanosensory and motor center (AMMC) and the gnathal ganglion (GNG). So far, no PN subtypes confined to the dmALT have been reported. This newly described neuron was therefore classified as Pdm_a (LPOG) . One LPOG output neuron projected in the antennal suboesophageal tract (AST; Figs. 4l, 6e). The small cell body of the neuron (approximately 4.5 μm in diameter) was localized outside the AL, in the dorsolateral cellbody rind in the ipsilateral GNG. The neuron projected ventrally, side by side with the sensory bundle from the www.nature.com/scientificreports/ LPO. Then it turned laterally and sent terminal projections into the ventral area of the AMMC and VLP. This newly described neuron was classified as Past_a (LPOG) .

Discussion
Here, we present new anatomical data on the central pathway processing CO 2 information in the noctuid moth, H. armigera. Totally, 15 individual AL PNs originating from the LPOG were stained. This morphologically heterogenous neuron assembly included nine subtypes, most of which were not previously described. A considerable amount of PNs, i.e. ten, passed along the up to now poorly described transverse ALT. The typical medial-tract neuron innervating the Ca before terminating in the LH, was not found. In order to estimate the real number of LPOG output neurons, we combined the results from the mass-staining and single-unit labeling experiments. The highest number of LPOG PNs observed in the anterogradely mass-stained preparations were five (Fig. 3a), including the three ipsilateral subtypes, Pt_a (LPOG) , Pm_e (LPOG) , and Pl_g (LPOG) . From the retrograde staining performed from the lateral protocerebrum, including nine stained somata linked to the LPOG in the contralateral AL (Fig. 3d), we can conclude that at least nine bilateral/contralateral PNs originate from the LPOG (possibly making up Pt_e (LPOG), Pt_f (LPOG) , and/or Pm_g (LPOG) subtypes). In addition, we stained three distinct PN subtypes not represented in any of the mass-stained preparations, i.e. Pt_d (LPOG) , Pdm_a (LPOG) , and Past_a (LPOG) (Fig. 4e, k,  l). Taken together, we assume there are at least 17 PNs originating from the LPOG. This number is comparable with the amount of CO 2 output neurons from the V-glomerulus in Drosophila, in which 12 PNs were imaged by using the photoactivatable green fluorescent protein technique 36 .
Notably, 12 of the 15 LPOG neurons presented here had overlapping projection terminals in the LH, which constitutes a main output target for PNs originating from the numerous ordinary glomeruli (see introduction). The comprehensive data on AL output neurons recently obtained from H. armigera-including both plant odor and pheromone neurons 25,26 , constitutes an excellent base for exploring putative interactions between the CO 2 pathway and these two olfactory sub-systems.

The CO 2 pathway and the sub-systems devoted to plant-odors and pheromones.
In Lepidoptera, CO 2 and plant odors are detected by sensory neurons localized on different organs. From here, the two categories of sensory neurons project to distinct sets of AL glomeruli, i.e., the LPOG and the ordinary glomeruli, respectively. As both stimulus inputs are relevant for detecting the quality of host plants (see introduction), it is likely that the two signal pathways interact in the central nervous system. Given that all LPOG PNs are uniglomerular, as those identified in this study, there seems to be a continuation of the segregated paths also at the second-order level, implying a system organized mainly according to a labeled-line principle 37 . On the other hand, the LPOG is not totally segregated from the other glomeruli but interconnected via numerous multiglomerular LNs innervating the AL globally. In a recent study of AL neurons in H. armigera, about 90% of the LNs were reported to arborize in the LPOG 26 . Even though a substantial proportion of these LNs had relatively weak innervations in the LPOG, this indicates that processing does occur across the glomerular categories. Like in other insects, the main proportion of LNs in heliothine moths is GABAergic 38 . Indeed, we can't rule out the possibility that LPOG PNs could be influenced by excitatory LNs [39][40][41][42] or by polysynaptic GABAergic inputs 43 . Anyway, considering the typical global pattern of LNs arborizing in the LPOG, we suggest that a modest degree of signal integration takes place across the sub-systems at this level.
The finding of uniglomerular LPOG neurons exclusively, forming a distinct output pathway for CO 2 information, as presented here, is in accordance with the labeled-line principle characterizing the two previously well-described olfactory sub-systems devoted to plant odors and pheromones, respectively 18,20,25,26,35 . Unlike the pheromone sub-system, having its distinct output area in the lateral protocerebrum, the CO 2 sub-system has axon terminals overlapping with the plant-odor PNs. As shown in Fig. 6b, reconstructions of two co-stained neurons, one from the LPOG and the other from an ordinary glomerulus, demonstrates the overlapping terminals in the LH. This overlap indicates that the LH is a significant region for integration of sensory inputs about plant odors and CO 2 . One putative downstream site for receiving information processed in the LH is the SMP 44 . Interestingly, the SMP is also directly innervated by at least two LPOG neuron subtypes identified here (Pt_d (LPOG) and Pdm_a (LPOG) ). Thus, to record response-patterns of SMP-neurons during stimulation with plant-odors and CO 2 constitutes an exciting subject for future studies. Taken together, we suggest that the LH is an essential site for integrating CO 2 and plant odor signals.
In contrast to the substantial overlap with parts of the plant odor sub-system, the LPOG projections had weak innervations in the main target areas of the male-specific MGC neurons, including the SLP and the superior intermediate protocerebrum (SIP) 25 . Although some of the LPOG PNs stained here, i.e. four neuron subtypes (Pt_a (LPOG) , Pt_d (LPOG) , Pm_e (LPOG) , and Pl_g (LPOG) ), had axonal terminals in the SLP, they targeted a distinct area localized more ventrally and lateral-posteriorly as compared to the MGC PNs 25,35,45 .
The finding that PNs carrying CO 2 signals interact more intensively with plant odor PNs than with pheromone PNs fits well with previous studies in Drosophila demonstrating extensive synaptic interactions in the LH between cholinergic PNs 46 . These cholinergic PNs include neurons responding to CO 2 47,48 . In addition, a population of plant/food odor PNs also target the same region 49 . Overall, the results presented here take us one step further towards understanding the neuronal architecture underlying moths' ability to measure fine fluctuations in CO 2 for determining the nutritional quality of relevant host plants 5,50 . CO 2 signaling is modestly involved in sensory learning. The calyces of the mushroom bodies are a neuropil for sensory integration and memory, important for experience-dependent reactions [51][52][53][54][55] . In contrast to the main olfactory pathway, which innervates the Ca extensively (Fig. 1) www.nature.com/scientificreports/ ably less terminals in the Ca than in the LH (Figs. 3a, 5,6). Actually, only one of the nine LPOG neuron subtypes, Pt_a (LPOG) , innervated the Ca. Its distributed and blebby terminals in the Ca correspond with the features of the typical uniglomerular plant-odor PNs and differ from the innervation pattern of pheromone PNs 18,25,35 . In the fruit fly, PNs tuned to CO 2 , which are connected with the V-glomerulus, also innervate the Ca modestly; the only neuron type targeting this neuropil is a bilateral lateral-tract neuron 36 . The general projection pattern of LPOG neurons bypassing the Ca suggests that the CO 2 signaling system is minimally involved in sensory learning, particularly olfactory learning. The apparent trivial role of CO 2 input for memory suits well with the fact that this gas is ubiquitous, serving as a minor compartment of the Earth's atmosphere. Like the CO 2 PNs, other AL output neurons tuned to universal cues, such as temperature and humidity, also display little or no connection with Ca. For instance, both in the fruit fly and cockroach, PNs carrying thermo-and hygro-sensory signals have modest or no innervation in the Ca 32,56 .
Functional implications of CO 2 signaling in the other brain areas. In addition to targeting the LH, the CO 2 PNs identified here also projected to various other neuropils, such as VLP, PLP, and SMP (see Fig. 4 and Table 1). In fruit flies, output neurons from the V-glomerulus, which corresponds with the LPOG, are reported to terminate in similar neuropil regions 36 . One major difference between the CO 2 systems in moth and fruit fly is that seemingly all LPOG neurons in moth are uniglomerular whereas the majority of those connected with the V-glomerulus in the fly is multiglomerular. Whether this is related to the peripheral organization in flies including CO 2 sensory neurons co-located with olfactory neurons on the antenna, is an open question. Regardless of dendritic arrangement in the AL, however, the second-order pathways for CO 2 signaling in the two insect groups are rather comparable.
Given these similarities across two insect model species, it is encouraging to speculate about the functional roles of these neuropils. The VLP, which is one of the major innervation areas of olfactory PNs 27 , is also targeted by neurons carrying information about vision 57 , audition 58 , and temperature and humidity 32 . We found that the VLP region is targeted by two transverse-tract LPOG subtype neurons, Pt_a (LPOG) and Pt_f (LPOG) , and an antennal subesophageal tract neuron, Past_a (LPOG) . Thus, the data presented here, coincides with the previous studies demonstrating the involvement of the VLP in integration of multimodal information. The PLP, which also constitutes an essential target area for olfactory neurons 27 , is reported to receive visual information both in fruit flies 32,57 and in the silk moth 59 . In the data presented here, we found at least four LPOG neuron subtypes (Pt_a (LPOG) , Pt_e (LPOG) , Pm_e (LPOG) , and Pl_g (LPOG) ) targeting the PLP region. Taken together, the previously considered integration areas for multimodal processing, the VLP and PLP, receive input about CO 2 as well.

Diverse transverse-tract neurons indicate their multifunctional roles.
In H. armigera, the transverse-tract neurons originate frequently from glomeruli localized in the ventral and posterior part of the AL 26 . We noticed that the majority of stained LPOG neurons, i.e., 10 of 15, were confined to this ALT. They are further classified into four morphological subtypes. Two subtypes, Pt_a (LPOG) and Pt_e (LPOG) , were previously described in M. sexta 34 -by then, they were classified as an inner-tract PIa(G) neuron and a protocerebral neuron, respectively. Unlike the classic ALTs (including the medial, lateral, and mediolateral ALT), the transverse ALT is known to be formed by a relatively restricted number of neurons displaying diverse morphologies 27,28 . This implies that the target areas of the transverse-tract neurons are less confined as compared to neurons in other tracts. A previous study on the fruit fly reported transverse-tract neurons originating from two posteriorly located glomeruli conveying information about temperature and humidity 32 . Notably, these non-odor responding neurons in the fly are morphologically comparable to some transverse-tract PNs found in H. armigera. Generally, the restricted number of neurons confined to the transverse ALT and their high degree of heterogeneity indicate that, in contrast to the main tracts, the tALT is formed by a few functionally unique PNs.

Conclusion
The anatomical data presented here, including visualization of individually labeled LPOG PNs forming the second-order pathway for CO 2 signals in the moth brain, contribute to improve our general understanding of parallel olfactory systems in insects. Besides, the high-resolution confocal data of individual neurons and neuron populations form the basis for future experiments exploring both anatomical and physiological characteristics of the central pathways involved in processing input about external fluctuations in CO 2 .

Materials and methods
Insects and preparation. Male and female Helicoverpa armigera pupae (Lepidoptera; Noctuidae, Heliothinae), obtained from Keyun Bio-pesticides (Henan, China), were allowed to eclose in climate chambers (Refritherm 200 and 6E, Struers-Kebolab, Albertsund, Denmark, or Binder KBF 720, Tuttlingen, Germany) at 24 °C and 70% air humidity on a 14:10 h light/dark cycle (lights on at 18:00). After emergence, the moths were supplied a 10% sucrose solution. The moths were 1-4 days old when the experiments were performed. According to Norwegian law of animal welfare, there are no restrictions regarding experimental use of Lepidoptera.
Preparation of the insect has been described in detail elsewhere 13,25 . Briefly, the moth was restrained inside a plastic tube with the head exposed and then immobilized with dental wax (Kerr Corporation, Romulus, MI, USA). The brain was exposed by opening the head capsule and removing the intracranial muscles and tracheas. The exposed brain was continuously supplied with Ringer's solution (in mM: 150 NaCl, 3 CaCl 2 , 3 KCl, 25 sucrose, and 10 N-tris (hydroxymethyl)-methyl-2-amino-ethanesulfonic acid, pH 6.9). www.nature.com/scientificreports/ Iontophoretic staining of individual PNs originating from the LPOG. The procedure of iontophoretic staining of AL PNs was performed as previously described 25,27,35 . Sharp electrodes were made by pulling quartz capillaries (OD: 1 mm, ID: 0.5 mm, with filament; Sutter instrument) on a horizontal puller (P-2000, Sutter instruments, CA, United States). The tip of the electrode was filled either with 4% biotinylated dextranconjugated tetramethylrhodamine (3000 mw, micro-Ruby, Molecular Probes) in 0.2 M potassium acetate (KAc) or 4% Alexa Fluor 488 dextran (10,000 mw, Molecular Probes) in distilled water. A chloridized silver wire placed in the eye served as a reference electrode. The recording electrode having a resistance of 200-300 MΩ, was backfilled with 0.2 M KAc. Iontophoretic staining of single neurons was conducted in two ways. (i) We carefully inserted the electrode into the LPOG region from the dorsal AL area (n = 10) via a micromanipulator (Leica, Wetzlar, Germany). Contact with an LPOG-neuron was confirmed by a response when applying a CO 2 puff (Fig. 3a). (ii) We pre-stained the LPO sensory neurons with micro-Ruby 48 h prior to the iontophoretic staining of LPOG output neurons. With clear visualization of the LPOG under a Zeiss stereo discovery V12 microscope equipped with epifluorescence, we inserted the electrode into the LPOG from a frontal position (n = 5). Each neuron was stained by applying 200 ms depolarizing current pulses of 2-3 nA, at 1 Hz, for 10 min. After labeling, the insects were kept overnight at 4 °C in dark to allow anterograde axonal transportation of the dye. Then the brain was dissected from the head capsule and 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 °C before it was 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).
Mass staining of output neurons originating from the LPOG. Mass staining of LPOG projection neurons was conducted in 8 male moths. Here, we applied the fluorescent dye into the relevant AL region via a sharp electrode with low resistance (~ 40 MΩ). Borosilicate capillaries (OD: 1 mm, ID: 0.5 mm, with filament 0.13 mm; Hilgenberg GmbH, Germany) were pulled on a horizontal puller (P-97, Sutter instruments, CA, United States). The tip of the glass electrode was filled with 4% micro-Ruby solution. The staining site was identified by testing the local field potentials during CO 2 stimulation (Fig. 3a). After that, the region was mass stained by applying a series of depolarizing current pulses of 50 nA at 1 Hz for 30 min. In addition to the anterograde mass staining, two types of retrograde labeling experiments were performed including application of dye crystals into the Ca and the LH, respectively. The dissection and dehydration were conducted as described above.
Stimulation delivery. During the electrophysiological recording, a pulse of CO 2 was delivered by a stimulation system including two parallel paths, one carrying a continuous airstream and the other the 400 ms CO 2 stimulus (100%, AGA AS, Oslo, Norway). A solenoid valve system (General Valve Corp.) regulated the application of the stimulus. To identify the neuron responses, application of the CO 2 stimulus and the control (fresh air) was repeated three times.
Confocal microscopy. Brains with successfully stained neurons were imaged using a confocal laser scanning microscope (LSM 800, Zeiss, Jena, Germany), equipped with C-Apochromat 10x/0.45 water objective, C-Apochromat 10x/0.3 air objective, and Plan-Neofluar 20x/0.5 air objective. Brains stained with micro-Ruby were scanned with a HeNe laser at 561 nm, and the emitted light was filtered through a 560-600 nm band pass filter. Preparations stained with Alexa Flour 488 were scanned with an Argon laser at 493 nm, including a 505-550 nm band pass filter. A synapsin-labeled preparation was carefully scanned utilizing two channels, in order to show that the neuropil structures visualized by the auto-fluorescent signal of endogenous fluorophores was in great agreement with the synapsin signal (Supplementary, Fig. S3). Thus, relevant structures in the brain containing the stained neurons were visualized by imaging the auto-fluorescence. Since many auto-fluorescent molecules in the tissue are excited at 493 nm, an Argon laser at 493 nm in combination with a 505-550 nm band pass filter was used. For all confocal scans, serial optical sections with a resolution of 1024 × 1024 pixels were obtained at 1.5-9 µm intervals through the entire depth of brain. The confocal images shown in this study were edited in ZEN 2.3 (blue edition, Carl Zeiss Microscopy GmbH, Jena, Germany). Brightness and contrast of the images were adjusted in Photoshop, version 21.
Reconstruction and registration of neurons into the representative brain. The 3D reconstruction of relevant neuropil structures was based on the previously conducted confocal stacks for the representative brain 25 . We used the nomenclature established by Ito, et al. 60 with the exceptions of the VLP. Our definition of the VLP, included the anterior and posterior VLP, along with the wedge. Furthermore, the superior intermediate protocerebrum (SIP) located adjacent to the SMP, was not specifically indicated in this study since none of the identified neurons projected to the SIP. Individually labeled neurons and the surrounding brain neuropils were manually reconstructed from consecutive confocal sections by means of the visualization software AMIRA 5.3; the neurons were reconstructed by using the skeleton module of the software 61,62 and the brain structures by using the segmentation editor. Thus, each neuron was traced so that a surface model built by cylinders of distinct lengths and thicknesses was created. The relevant neuropils were reconstructed based on the autofluorescence signals, as described above. To compensate for the refraction indexes, the z-axis dimension of the neuron and brain structures was multiplied by a factor of 1.16 for the water lens objective and 1.54 for the dry lens. Manual registration of individual neurons into a representative brain from Chu et al. 25 followed the same procedure as described in Ian et al. 27 .
To provide a framework for determining the target areas of the LPOG PNs, we reconstructed nine new protocerebral regions from the confocal image stack previously utilized for creating the representative brain model 25  www.nature.com/scientificreports/ utilizing the segmentation editor of AMIRA. Unspecified neuropil regions have previously been characterized in the lepidopteran brain by Ian et al. 27 and Heinze and Reppert 63 .
Image intensity processing and statistical analysis. As retrograded staining from the Ca labeled dendrites of all AL PNs linked with this neuropil, we could investigate putative differences in connection pattern between the LPOG and a typical ordinary glomerulus by measuring the glomerular fluorescence intensities.
Here, we compared the fluorescence intensities in the LPOG and the neighboring glomerulus, G71, in six moths. We selected four confocal sections including all parts of each glomerulus. The fluorescence intensity within each of these two glomeruli was quantified by using Image J, respectively (https ://image j.net). Paired sample t test was performed to compare the mean intensities. All probabilities given are two-tailed. The Statistical package for the social sciences (SPSS), version 25, was used for statistical analysis.