Antennal‐lobe neurons in the moth Helicoverpa armigera: Morphological features of projection neurons, local interneurons, and centrifugal neurons

Abstract The relatively large primary olfactory center of the insect brain, the antennal lobe (AL), contains several heterogeneous neuronal types. These include projection neurons (PNs), providing olfactory information to higher‐order neuropils via parallel pathways, and local interneurons (LNs), which provide lateral processing within the AL. In addition, various types of centrifugal neurons (CNs) offer top‐down modulation onto the other AL neurons. By performing iontophoretic intracellular staining, we collected a large number of AL neurons in the moth, Helicoverpa armigera, to examine the distinct morphological features of PNs, LNs, and CNs. We characterize 190 AL neurons. These were allocated to 25 distinct neuronal types or sub‐types, which were reconstructed and placed into a reference brain. In addition to six PN types comprising 15 sub‐types, three LN and seven CN types were identified. High‐resolution confocal images allowed us to analyze AL innervations of the various reported neurons, which demonstrated that all PNs innervating ventroposterior glomeruli contact a protocerebral neuropil rarely targeted by other PNs, that is the posteriorlateral protocerebrum. We also discuss the functional roles of the distinct CNs, which included several previously uncharacterized types, likely involved in computations spanning from multisensory processing to olfactory feedback signalization into the AL.

usually positioned outside the AL.
The PNs constitute a heterogeneous neuronal group carrying odor information from the AL to higher brain centers via several parallel antennal-lobe tracts (ALTs). Altogether, moths have 850 PNs restricted to six ALTs in each brain hemisphere, that is the medial, mediolateral, lateral, transverse, dorsomedial, and dorsal ALT (m-, mll-, t-dm-, and d-ALT, respectively; see Figure 1). As the majority of the PNs follow the first three ALTs, they are considered the main tracts (Homberg et al., 1988;Ian, Berg, Lillevoll, & Berg, 2016). PNs principally target two main areas-the calyces of the mushroom bodies and the lateral protocerebrum, including the lateral horn (LH; F I G U R E 1 Schematic view of the parallel antennal-lobe tracts (ALTs) formed by projection neurons (PNs). The PNs of the antennal lobe (AL), following one of six distinct ALTs, vary in axonal projection patterns, glomerular arborizations, and soma location. Here, the thickness of each ALT roughly relates to its estimated number of PNs. The medial-tract (mALT) PNs target the calyces (Ca) and lateral horn (LH), and have somata in all three AL cell clusters, that is the anterior, lateral, and medial clusters (AC, LC, and MC, respectively). The lateraltract (lALT) PNs have more heterogeneous projection patterns, including output areas in the ventrolateral protocerebrum (VLP), LH, and the column of the superior intermediate protocerebrum (SIP). The mediolateral-tract (mlALT) PNs typically innervate the LH and the superior lateral protocerebrum (SLP), but may also target other neuropils, for example, the posteriorlateral protocerebrum (PLP). Transverse-tract (tALT) PNs have varying target regions, including the Ca, LH, and PLP. PNs of the lALT, mlALT and tALT all have somata in the LC. The dorsal-tract (dALT) PNs have their somata in the contralateral MC, and target many protocerebral neuropils, including the VLP, LH, PLP, SLP, and the superior medial protocerebrum (SMP). Bilateral dorsomedial (dmALT) PNs have somata in the subesophageal zone (SEZ), innervate one glomerulus in the AL of each hemisphere, and project bilaterally to the Ca, PLP and LH. Glomerular arborization pattern differs between ALTs; the PNs may be uniglomerular (mostly mALT and dmALT), oligoglomerular (tALT, lALT, and mlALT), or arborizing in most glomeruli (dALT and some PNs from the lALT, and mlALT). L, lateral; P, posterior [Color figure can be viewed at wileyonlinelibrary.com] Homberg et al., 1988;Ian, Zhao, Lande, & Berg, 2016). So far, uniglomerular PNs in the mALT, which innervate the two former protocerebral areas, are most frequently reported. The PNs following the mALT have their somata located in one of the three cell body clusters, whereas PNs confined to the other main tracts have their somata in the LC (Homberg et al., 1988). The dendritic and axonal arborization patterns of individual PNs within any given ALT are quite diverse (see Table 1 for summary of previous reports in moths).
Actually, this diversity of PN types is rather common in insects. Even the small fruit fly, having only 350 PNs (Bates et al., 2020), possesses at least 15 different PN sub-types (Tanaka, Endo, & Ito, 2012).
In comparison with PNs and LNs, the CNs pose a minor category.
To extend our understanding of how odor signals are processed in the AL and how this information is conveyed to higher brain centers, we have elucidated the structure of the basic coding element in this system, that is the individual neuron. By using iontophoretic staining, combined with confocal microscopy, we have systematically gathered anatomical data from individual neurons in the heliothine moth, H. armigera. The total collection of 190 AL neurons identified and presented here, covers all three categories-PNs, LNs, and CNs.
In addition to previously reported neuron types, we present several novel ones. Altogether, the data presented here provide a comprehensive and detailed overview of the neurons forming the primary olfactory processing center.

| Insect preparation
Pupae of H. armigera were obtained from Keyun Bio-pesticides (Henan, China). All neurons were sampled from males, unless otherwise noted. The insects were kept in climate chambers at 23 C and 70% humidity, with reversed day-night cycle (light-dark 14:10 hr).
After emergence, the moths were provided a 10% sucrose solution.
According to Norwegian law of animal welfare there are no restrictions regarding experimental use of Lepidoptera.
The moth was placed in a plastic tube, where the protruding head was immobilized using a layer of dental wax (Kerr Corporation, Romulus, MI). To access the brain, a part of cuticle and antennal muscles were carefully removed. Ringer solution (in mM: 150 NaCl, 3 CaCl2, 2 KCl, 25 sucrose, and 10 N-tris [hydroxymethyl]-methyl-2-aminoethanesulfonic acid, pH 6.9) was applied continuously during the experiments.

| Intracellular staining and confocal microscopy
The staining procedure was performed as previously described Ian, Zhao, et al., 2016;KC et al., 2020

| Reconstruction
Digital reconstruction of individual neurons, acquired by confocal scanning, was used to visualize individual neurons. In some cases, reconstruction also aided visualization of weakly stained filaments. The individual neuron was traced manually by using the SkeletonTree plugin (Evers, Schmitt, Sibila, & Duch, 2005;Schmitt, Evers, Duch, Scholz, & Obermayer, 2004) in 3D reconstruction software Amira 5.3 (Visualization Science Group). To compensate for refraction indexes, the z-axis dimension of the brain was multiplied by a factor of 1.16 for the water objective and 1.54 for the air lens. Reconstructed neurons were manually transformed into the 3D representative brain of H. armigera  in accordance with their original projection pattern.

| Nomenclature and neuronal classification
The nomenclature was based on the work of the Insect Brain Name Working Group (Ito et al., 2014), which used the fruit fly as a model organism. Determining the position of distinct neuropils was based upon inspection of 3D models of standardized brains from the fruit fly (Ito et al., 2014) and the monarch butterfly, Danaus plexippus (Heinze, 2016), but adjusted in accordance with landmarks in the moth brain. The classification of PNs also built on previous naming systems used in the moth species, M. sexta (Homberg et al., 1988) and Heliothis virescens (Ian, Zhao, et al., 2016). We excluded PNs classified exclusively as MGC-or LPOG-neurons, these will be presented in separate studies. In addition, electrophysiological data will be presented in an upcoming article.
For multiglomerular neurons, classification of dendritic morphology was based on the nomenclature for LNs in the silkmoth, Bombyx mori (Seki & Kanzaki, 2008). Identification of specific AL glomeruli and nerve bundles followed the arrangement specified for H. armigera , except for the ventroposteriorly located glomeruli, the VPGs. Our new definition of a glomerular group called the VPGs is based on the finding of an assembly of heteromorphic glomerular units that were frequently interconnected by oligoglomerular PNs arborizing primarily or exclusively in these specific glomeruli. Furthermore, these glomeruli were weakly labeled by previous OSN mass dye fills . The VPGs are proposed to include at least three glomeruli located posteriorly to the LPOG, that is G71-G73, as well as G64-G69, which form a horseshoe-shaped cluster of undersized glomeruli ventral to the AL hub.
In some cases, distinguishing between CNs and PNs was complicated due to combinations of smooth and varicose processes in the same brain regions. Such structures are associated with post-and presynaptic sites, respectively (Cardona et al., 2010). In these instances, soma location was taken into consideration, in the sense that a soma found in the protocerebrum would indicate that the neuron was centrifugal.

| RESULTS
We present a collection of individually labeled neurons consisting of all three AL neuron categories, that is PNs, LNs, and CNs.

| Antennal-lobe projection neurons
A PN connects the AL and other brain regions via one of six ALTs. We We classified these 109 PNs into different types according to the ALT the axon projected through and divided each PN type into distinct sub-types based primarily on axonal projection patterns in the protocerebrum. For clarity, for the Pm_a neurons, P indicates that the category is PN, m applies to the medial-tract type, while a refers to the sub-type. This naming system is an adaptation of the terms introduced by Homberg et al. (1988).

| Projection neurons in the medial antennallobe tract
In total, 67 mALT neurons were stained across 57 preparations. As many as 61 of these PNs were classified as Pm_a sub-type, that is largely homogenous uniglomerular PNs targeting the calyces and LH ( Figure 2a; Homberg et al., 1988;Ian, Zhao, et al., 2016). These PNs had somata in all three AL cell clusters, with 38% in the LC, 44% in the MC, and 18% in the AC. All reported Pm_a neurons were restricted to OGs or PCx glomeruli, and had comparable protocerebral innervations. However, some PCx PNs innervated more dorsomedial parts of the LH than the OG PNs (e.g., PN3; Figure 7). Unlike most previous reports, we found that a few Pm_a neurons were not uniglomerular. Four Pm_a neurons innervated two OGs-filling one OG densely, and the outer layer of another OG more sparsely ( Figure 2b). Furthermore, another two co-stained Pm_a neurons (PN15; Figure 7) collectively innervated six OGs.
Six neurons of the Pm_e sub-type were labeled. As previously reported by Ian, Zhao, et al. (2016), these PNs project terminal branches OGs. Some of these PNs were co-stained with LNs, therefore only traceable dendrites were considered as part of the Pm_e neurons.

| Projection neurons in the lateral antennallobe tract
Altogether 18 PNs in the lALT were stained across 17 preparations; these consisted of four distinct sub-types. All lALT PNs had somata in the LC, while dendritic and axonal innervations differed across and within most sub-types, confirming the previously reported heterogeneity of neurons in this tract (Homberg et al., 1988;Ian, Zhao, et al., 2016).
Two Pl_b neurons were labeled, both terminated primarily in the VLP. One PN had pronounced club-like terminals (Figure 3a), as reported previously in H. virescens (Ian, Zhao, et al., 2016) and M. sexta (Homberg et al., 1988). The two neurons' dendritic innervation patterns differed, as one was uniglomerular while the other was oligoglomerular.

| Projection neurons in the mediolateral antennal-lobe tract
Thirteen mlALT PNs were stained. All passed along the initial part of the mALT, then bent laterally at the anterior edge of the central body, running towards the lateral protocerebrum. All Pml neurons had their somata in the LC, and with one exception, all had multiglomerular dendritic innervations, which were quite consistent within sub-types.
In addition to the previously reported Pml_b (Ian, Zhao, et al., 2016), we present two additional sub-types.
The Pml_b sub-type included four stained PNs sending their axon to the LH and surrounding neuropils, including the PLP, SLP and the anterior base of the calyces (bCa; Figure 4a). In addition, PN75 and PN77 innervated regions along the border of the SCL and the superior medial protocerebrum (SMP). All Pml_b neurons were oligoglomerular, with dendrites limited to at most four heterogeneously innervated VPGs or PCx glomeruli.
Five labeled mlALT PNs were found to constitute a novel subtype, which we termed Pml_c. A prominent Y-shaped bifurcation of the axon, located ventrally to the pedunculus, typified the Pml_c neurons. From here, the main fiber targeted the LH and PLP, before turning dorsomedially to the SLP and bCa. In addition, a thin sub-branch

| Projection neurons in the transverse antennal-lobe tract
Six tALT neurons were labeled, including one bilateral PN. These PNs formed four sub-types, all following the mALT before bending laterally at the posterior border of the central body, between the turning points of the m-and ml-ALTs. The axons of all unilateral tALT neurons bifurcated when leaving the mALT, and both sub-branches often converged onto the same neuropil. The PNs we classify as Pt_a and Pt_d have previously been reported as mALT PNs, that is as Pm_c and Pm_b sub-types, respectively (Homberg et al., 1988;Rø et al., 2007).
The most prevalent tALT sub-type, which we named Pt_a (

| Antennal-lobe centrifugal neurons
Fourteen labeled AL CNs formed seven distinct types (Table 4;  (AOTU). The two neurites then ran ventroanteriorly, entering the AL ventromedially to the mALT. In each AL, most glomeruli were innervated by blebby terminals, however the size and density of these terminals appeared to be bilaterally asymmetric.
The second bilateral CN type was the CSDn (Figure 10b), which has been examined in many insect orders (reviewed by Dacks, Christensen, & Hildebrand, 2006).
Sparse --2 2 -Note: Normal innervation within a sub-system is here defined as most or all glomeruli within the specific sub-system receiving processes that are comparable to that of the ordinary glomeruli (OGs through the antenno-subesophageal tract (AST). All contralateral glomeruli were densely innervated, while only a few OGs were innervated in the ipsilateral AL. We termed this CN type, Cast.

| Unilateral centrifugal neurons
The unilateral CNs consisted of four types: a novel medial-tract CN called Cm, two novel lateral-tract CN types, Cl_a and Cl_b, and one CN type resembling a previously reported multisensory CN (Zhao et al., 2013). The last-mentioned type was termed Csfs, as its protocerebral neurites appeared to follow what Ito et al. (2014) calls the superior fiber system (SFS).
The Cm type was stained in three preparations (Figure 11a

| DISCUSSION
In this work, we present a comprehensive collection of individual AL neurons from the moth brain-including all three categories. Except for a previous report from Tanaka et al. (2012), who mapped AL-neurons in the fruit fly using genetically modified strains, no corresponding overview of such neurons has been described in any insect species. We found that the 190 stained neurons consisted of 25 distinct neuronal types or sub-types, many of them novel. Generally, all AL neuron categories consist of various heteromorphic types, suggesting complex olfactory processing already at the first synaptic level of the sensory pathway.
In comparison with mammals, the number of parallel tracts is substantially higher in insects. One reason for this discrepancy may be the fact that insects possess a high proportion of multiglomerular PNs, whereas mammals have only uniglomerular PNs (Buck & Bargmann, 2013).
In moths, most uniglomerular PNs, assumed to be essential to odor discrimination, are confined to the mALT (Hansson et al., 1994;Homberg et al., 1988;Ian, Zhao, et al., 2016;Kanzaki et al., 1989). In this study, about half of all labeled PNs were morphologically homogenous uniglomerular Pm_a neurons, which might be compared with mammalian mitral and tufted cells. Both mammalian and insect uniglomerular PNs innervate higher centers involved in identification of odor quality, that is the piriform cortex (Gottfried, 2010) and calyces (Galizia, 2014), respectively. Our data demonstrate that tracts other than the mALT are formed mainly by axons of multiglomerular PNs, with the exception of the dmALT consisting of bilaterally uniglomerular PNs (Figure 6a). Generally, the multiglomerular PNs do not target the calyces. One central difference between the roles of uni-and multiglomerular PNs may relate to combination selectivity. We found that the glomerular innervations of the latter PNs were often distant from the putative spike initiating proximal parts of the dendritic tree (Christensen, D'Alessandro, Lega, & Hildebrand, 2001;Gouwens & Wilson, 2009). As electrophysiological signals attenuate while traveling downstream towards a spike initiation zone (Rall & Rinzel, 1973), these PNs should be most responsive to specific stimuli that activate several of the contacted glomeruli. This is similar to the combination selectivity of pyramidal cells in the mammalian piriform cortex (Kumar et al., 2018), even though it occurs at a lower processing level. While the combination selectivity of a single neuron in mammals is random, it may be hardwired in the multiglomerular PNs of the moth, as indicated by consistent AL innervations of specific PN sub-types across individual moths.
The output targets vary across PN types, although the LH is innervated by most PNs. In line with former studies, all Pm_a neurons from ordinary glomeruli target the calyces and the LH exclusively (Homberg et al., 1988;Martin et al., 2011). The mostly multiglomerular lALT PNs, on the other hand, project primarily to more ventral neuropils such as the VLP (Figure 3; Homberg et al., 1988;Ian, Zhao, et al., 2016). The Pl_a sub-type targeting the column of the SIP is not represented here. All stained PNs of this sub-type innervated the MGC, and they were recently reported in Chu et al. (2020). The almost exclusively multiglomerular mlALT PNs all innervate the SLP (Figure 4), and most also target the LH. The three remaining tracts are, so far, rarely or incompletely described at the level of individual neurons. We found that PNs in the recently identified tALT Ian, Zhao, et al., 2016;Tanaka et al., 2012) are mostly oligoglomerular and project to the PLP, while other protocerebral targets vary across sub-types ( Figure 5). The labeled dALT PN had widespread axon terminals and connected to large parts of the AL (Figure 6b). As many dALT and mlALT PNs are GABAergic (Berg et al., 2009), they can provide inhibitory input to many protocerebral regions. Finally, the dmALT PNs resemble the mALT PNs in many ways, including innervation of the calyces and LH (Ian, Zhao, et al., 2016;Kanzaki et al., 1989;Rø et al., 2007). However, we found that these uniglomerular PNs only innervated one of the VPGs, had somata located outside the AL, and extended additional terminal branches into the PLP.

| Connections between distinct PN types/ sub-types and ventroposterior glomeruli
By investigating the morphological data from many individually stained PNs passing along different tracts, we could describe glomerular arborization patterns across PN types. An interesting observation concerns the VPGs (see definition in section 2.4). These heteromorphic glomeruli were previously found to be substantially weaker labeled by OSN mass dye fills than most other glomeruli , indicating that they receive relatively little antennal input. In addition, the LN branching patterns found within the VPGs in this study is distinct from that in the OGs, resembling how the MGC versus OGs arborizations differ. Taken together, this implies that the VPGs may pose its own functional AL sub-system. In total, 20 of the reported 109 PNs had dendrites completely or primarily confined to the VPGs. Notably, none of the 61 labeled Pm_a neurons innervated the VPGs.
Interestingly, all four bilateral dmALT PNs innervated both ALs with tightly packed dendrites in one of two glomeruli, G71 or G72, exclusively ( Figure 6a). Up to now, only a few incompletely labeled neurons in this thin fiber bundle, which is estimated to include ca. 16 axons in M. sexta (Homberg et al., 1988), has been described.
Notably, the two glomeruli are clustered together with the LPOG, which handles input about CO 2 rather than general odors.
Furthermore, some of the PNs connected with the VPGs appear to combine AL signals with inputs from the AMMC and SEZ. This where seven VPGs process thermo-and hygro-sensory inputs, we find a cross-species pattern of VPG PNs terminating in the ventroposterior LH and PLP (Frank et al., 2015;Marin et al., 2020). In addition, temperatureand humidity-processing PNs in the cockroach have corresponding morphological traits (Nishino et al., 2003). The relevant glomeruli in these insects are generally heteromorphic, like the moth VPGs, and unlike the spheroidal OGs. PNs innervating corresponding glomeruli in other insects consist of many types, several of which are similar to those reported here.
For example, findings from the cockroach include PNs resembling both the Pl_c and Pm_e sub-types (Figures 7 & 8 in Nishino et al., 2003). In the fruit fly, tALT PNs and an analogue of the dmALT PNs innervate the LNs. An MGC-AllGs LN is presented in a1-a6. A maximum intensity projection of the confocal stacks covering the entire AL can be seen in a1, while the following images demonstrate selected regions of the AL, with specific glomerular sub-systems indicated by dashed lines. The labial palp-pit organ glomerulus (LPOG) was sparsely innervated (a2), while the ventroposterior glomeruli (VPGs; a3) included more processes. As in all MGC-AllGs LNs, the ordinary glomeruli (OGs; a4) were substantially innervated. This particular LN also innervated the macroglomerular complex and posterior complex (MGC and PCx; dashed and dotted lines in a5, respectively). In a6, a reconstruction of this LN is presented.

| The ventrolateral protocerebrum-A putative area for multimodal processing
The VLP was innervated by all Pm_e and Pml_d neurons, most lALT sub-types, and by the unique dALT PN. Altogether, the different PN types and sub-types projecting to the VLP originate from diverse AL sub-systems (see Supplemental Table 1), and include PNs innervating the SEZ and AMMC (Pm_e and Pl_d), thereby having the potential to function as multimodal integrators (Homberg et al., 1988;Tanaka et al., 2012). Previously, the VLP of heliothine moths was reported to be innervated by auditory ascending neurons (Pfuhl, Zhao, Ian, Surlykke, & Berg, 2014) and gustatory interneurons (Kvello, Løfaldli, Rybak, Menzel, & Mustaparta, 2009). The VLP may therefore serve as a multimodal integration center. In the fruit fly, the ventral LH and PLP, both localized adjacent to the VLP, are convergence sites for F I G U R E 1 1 Legend on next page. multimodal inputs, including information about aversive odors , mechanosensation, audition (Dolan et al., 2019), thermoand hygro-sensation (Frank et al., 2015;Frank et al., 2017;Marin et al., 2020), and gustation (Kim, Kirkhart, & Scott, 2017).

| Information processing in CNs
Contrary to the PNs, the CNs modulate AL activity based on input in other brain regions. However, CN processes in the AL are not only presynaptic, they may be postsynaptic as well (Sun, Tolbert, & Hildebrand, 1993;Zhang & Gaudry, 2016), which is also the case for OSNs, LNs, and PNs (Bates et al., 2020;Rybak et al., 2016;Shimizu & Stopfer, 2017;Sun, Tolbert, & Hildebrand, 1997;Tabuchi et al., 2015). This bidirectional processing complicates classification

| CNs potentially playing a role in olfactory feedback modulation
Several of the CNs identified here, may play a role as olfactory feedback neurons, based on their protocerebral dendrites being localized in odor-centers. The three labeled DAAR neurons, for example, arborized in the SMP, SIP, crepine, and SLP ( Figure 10a). In B. mori, a neuron resembling the DAAR neuron, albeit with unilateral AL innervation, was reported to respond to odor stimulation (Namiki, Iwabuchi, Kono, & Kanzaki, 2014). Indeed, olfactory input could be transmitted to the DAAR neuron`s protocerebral branches both via second order PNs (e.g., Pml_d) and via third order olfactory neurons which innervate relevant regions in both moths and fruit flies (Dolan et al., 2019;Namiki et al., 2014). The DAAR neurons labeled here differed from the otherwise comparable DAAR CNs reported in M. sexta by innervating a visual neuropil, the AOTU, rather than the LH (Dacks et al., 2012). Three additional CN types, Cl_a, Cl_b, and Cm, all had dendrites in higher olfactory regions, such as the LH and VLP, suggesting that they may act in odor feedback modulation as well.
Interestingly, some of the putative feedback CNs mentioned above may also contribute in local lateral processing within the AL. For instance, the Cl_b neuron ( Figure 11d) had fine arborizations in medial OGs, while the dorsal AL and the AL hub received extraglomerular varicose processes, which implies post-and presynaptic sites, respectively. As the AL hub contains neurites of most AL neurons (Homberg et al., 1988),

| CN morphology indicates memory-guided AL influence
The Cm neurons had dendrites in the calyces, medial lobes, and various other protocerebral neuropils, and were thus the only CNs with extensive mushroom body innervations (Figure 11a-b). Given the involvement of these structures in establishment of odor-memory (reviewed in Heisenberg, 2003;Stopfer, 2014), the Cm neurons might offer memory-guided information to the AL. . In addition, the medial lobes (ML), the adjacent crepine (CRE), and the calyces (Ca) were innervated. An axon projected to the antennal lobe (AL) through the medial antennal lobe tract (mALT). All AL glomeruli were filled, but a cluster of anterior ordinary glomeruli (OGs) received particularly large terminals. (a3) Dorsofrontal image displaying branches following both the mALT and the lateral ALT (lALT; black arrowhead). (b1-b3) another Cm neuron, sampled from a female moth. The two Cm neurons were largely similar, including the sub-branch in the lALT (black arrowhead in b3). (c1-c3) a female moth's Cl_a neuron, with its soma in the PCBR. The dendrites innervated the VLP, LH, PLP, SIP, and the base of the calyces (bCa). The axon followed the lALT to the AL, where approximately 10 dorsoposterior glomeruli were innervated. The neurite then projected ventrally into the antennal mechanosensory and motor center (AMMC) and the subesophageal zone (SEZ; see sagittal view in c2). (d1, d2) the Cl_b neuron, which had its soma in the anterior cell body rind. Dendrites innervated the VLP, LH, PLP and SLP. The axon followed the lALT into the AL, where medial OGs were innervated with small processes, while another branch sent large varicosities into the antennal lobe hub (ALH; d3). In addition, the latter branch terminated in the dorsal AL, primarily in extraglomerular spaces (d4). (e1, e2) a Csfs neuron, following the superior fiber system (SFS). The CN's soma was in the cell body rind along the midline (MCBR), contralateral to all innervations. The dendrites ran through the CRE, SIP, SLP, SCL, and SMP. The AL processes included innervation of one dorsoposterior OG (e3), along with branches in the ALH (e4). (f1, f2) dorsal confocal stacks from a mass staining of an AL demonstrated three labeled MCBR somata, along with several branches following the SFS. Two somata were located contralaterally (CL) to the labeled AL, while one was in the ipsilateral hemisphere (IL). All figures are in dorsal orientation, unless otherwise stated. D, dorsal; L, lateral; M, medial; P, posterior. Scale bars = 100 μm [Color figure can be viewed at wileyonlinelibrary.com] called anterior inferiormedial protocerebrum (Ito et al., 2014), is innervated by at least 12 types of mushroom body neurons (Tanaka, Tanimoto, & Ito, 2008). Provided that the crepine in moths also connects with the mushroom body, these CN types may also be involved in memory processes.

| The Cast neuron may be a mechanosensory peptidergic CN
As demonstrated, the Cast neuron had anatomical features indicative of bilaterally asymmetric AL modulation via antennal-mechanosensory input (Figure 10c). Mechanosensory responses have been described in AL neurons (Kanzaki et al., 1989;Zhao & Berg, 2010), but mediation of such responses could also arise directly from mechanosensory neurons on the antennae that terminate in the AL (Han, Hansson, & Anton, 2005).
The full morphology of the Cast neuron has not been previously reported, but a similar tachykinin-immunoreactive (TKir) CN with only contralateral innervations has been presented (Berg et al., 2007). That TKir CN could have been incompletely reconstructed, possibly due to immunolabeling of a mirror-symmetric CN interfering with neuritetracing. One recent report contradicts the presence of a TKir CN, claiming these neurons to bypass the AL on its ventroposterior side (Zhao, Xie, Berg, Schachtner, & Homberg, 2017), again based on immunolabeling. However, our intracellularly labeled Cast neuron does not only fit the relevant morphological description in Berg et al. (2007), but also corresponds to the confocal images of TKir neurons presented in Figure 1B-D of Zhao et al. (2017).

| On local interneuron diversity
The majority of labeled LNs, 60 of 67, was classified as MGC-AllGs type, innervating most or all glomeruli (see Figures 8 & 9). The fact that 30% of these LNs had substantially sparser innervations in the MGC than in the OGs, is in accordance with findings from M. sexta and B. mori (Matsumoto & Hildebrand, 1981;Seki & Kanzaki, 2008). This indicates a reduced role for LN-mediation in the MGC, yet MGC-PN activity is clearly mediated by GABA (Christensen, Waldrop, & Hildebrand, 1998;Lei, Christensen, & Hildebrand, 2002), which corresponds with the large proportion of GABAergic LNs found across moth species (Berg et al., 2009;Seki & Kanzaki, 2008). The MGC-AllGs LNs further included 26%, 20%, and 14% with sparse or no innervations in the LPOG, PCx, and VPGs, respectively. Moreover, all labeled oligoglomerular LNs primarily innervated the OGs. As such, OG processing appears to be the principal focus of the LNs.
We also report one novel LN type, that is the MGC-AllGs-IST type ( Figure 8e), with branches stretching to the AL isthmus. Neurons extending beyond the AL would usually be classified as PNs, however the neurites in question did not appear to be axonal, nor did these LNs sample inputs primarily beyond the AL, like the CNs do. As this LN type had processes in the AL isthmus, it may have dendro-axonic connections with lALT PNs, which also arborize here (for example, see PN61 in Figure 7; Homberg et al., 1988).

| CONCLUSION
Here, we present a comprehensive collection of AL neurons in the moth H. armigera. The extraordinary morphological diversity of these neurons indicates the advanced chemosensory processing taking place already at the initial synaptic level of the olfactory pathway. In addition to the scientific value of documenting the peculiarities and complexity within a small insect brain, the anatomical data may serve as a basis for future studies aiming at exploring physiological characteristics of the various neuron types.