3D atlas of cerebral neuropils with previously unknown demarcations in the honey bee brain

Honey bees (Apis mellifera) express remarkable social interactions and cognitive capabilities that have been studied extensively. In many cases, behavioral studies were accompanied by neurophysiological and neuroanatomical investigations. While most studies have focused on primary sensory neuropils, such as the optic lobes or antennal lobes, and major integration centers, such as the mushroom bodies or the central complex, many regions of the cerebrum (the central brain without the optic lobes) of the honey bee are only poorly explored so far, both anatomically and physiologically. To promote studies of these brain regions, we used anti‐synapsin immunolabeling and neuronal tract tracings followed by confocal imaging and 3D reconstructions to demarcate all neuropils in the honey bee cerebrum and close this gap at the anatomical level. We demarcated 35 neuropils and 25 fiber tracts in the honey bee cerebrum, most of which have counterparts in the fly (Drosophila melanogaster) and other insect species that have been investigated so far at this level of detail. We discuss the role of cerebral neuropils in multisensory integration in the insect brain, emphasize the importance of this brain atlas for comparative studies, and highlight specific architectural features of the honey bee cerebrum.


INTRODUCTION
The honey bee (Apis mellifera) has been studied extensively due to its elaborate sensory systems, sophisticated neuronal processing, remarkable behavioral plasticity, and cognitive capabilities (e.g., Brill et al., 2013;Frisch, 1993;Groh & Rössler, 2020;Menzel & Giurfa, 2001). This includes, for example, olfactory and visual processing, learning and memory capacities, communication skills, spatial orientation, and multisensory navigation. The general neuroanatomy of the honey bee brain has been pioneered starting in the 19th century (e.g., Dujardin, 1850;Kenyon, 1896;Mobbs, 1984;Vowles, 1955). More recently, increasing numbers of studies combined neuroanatomical analyses with molecular tracing, confocal imaging, and computer-based 3D reconstruction to map neuropils, individual neurons, and neuronal tracts in insect brains. Pioneering studies using 3D reconstructions from confocal images have been done in both Drosophila melanogaster and in the honey bee (Brandt et al., 2005;Rein et al., 2002). These were followed by 3D brain atlases obtained from other insect species and the introduction of a unified nomenclature for insect brain neuropils and www.insectbraindb.org) .
The first honey bee 3D brain atlas by Brandt et al. (2005) had been designed as an average-shape atlas reflecting the 3D outlines of major neuropils from a population of honey bee workers. The aim of this average brain atlas was that it can be used for adding individual neurons that have been physiologically recorded and morphologically traced in other individuals using morphological transfer via elastic registration (Brandt et al., 2005;Rybak et al., 2010). The average-shape atlas of the honey bee brain comprises major neuropils like the optic and antennal lobes, the mushroom bodies, the central body of the central complex, and major parts of the central brain as a single neuropil mass, termed protocerebral lobe in Brandt et al. (2005). Individual compartments of the cerebrum (comprising the fused proto-, deuto-, and tritocerebrum without the optic lobes), however, were not further subdivided in the average-shape atlas by Brandt et al. (2005). As these neuropils have less distinct boundaries, their demarcation remained unknown in the honey bee brain. This brain region had been termed "terra incognita" (Ito et al., 2014) and has first been mapped systematically in the fruit fly, Drosophila melanogaster, with the intent to provide a unified nomenclature of neuropils and major tract systems of the entire insect brain.
This individual brain atlas together with the unified nomenclature for synapse-rich neuropils by Ito et al. (2014) not only promoted neuronal circuit analyses in previously unknown neuropils in various parts of the cerebrum of Drosophila, but also opened a comparison of neuroanatomical features revealed by brain atlases from other insect species.
A precise anatomical map of cerebral neuropils is important to understand multisensory convergences and integration in the insect brain. For example, previous studies have identified neuropils in the cerebrum of the honey bee as target regions of mechanosensory afferents from the antennae and visual projections from the ocelli and optic lobes (e.g., Ai et al., 2007;Maronde, 1991). More recently, tracings of afferent projections from the Johnston's organ have been mapped into the individual 3D atlas of the Cataglyphis ant brain and suggest that the posterior slope and ventrolateral protocerebrum are involved in multisensory integration (Grob et al., 2021;Habenstein et al., 2020). The integration of input from different sensory modalities and their connection with premotor centers represent important prerequisites for understanding the neuronal control of navigation and cognitive capabilities (Currier & Nagel, 2020). As another example, a detailed 3D atlas of the honey bee cerebrum is required for a deeper analysis of target regions of output neurons from the mushroom bodies (MB), brain centers involved in sensory integration, learning, and memory. Previous studies had shown that mushroom body output neurons (MBON) populate large areas of the superior and lateral part of the protocerebrum (Rybak & Menzel, 1993). To properly map potential feedback circuits of MBONs with upstream olfactory or visual sensory pathways and their connections with downstream premotor pathways, the detailed analysis of cerebral neuropils and tract systems is a crucial requirement. In Drosophila and the bumble bee, neuropil atlases at the light microscopic level promoted follow-up studies employing highresolution electron microscopy-based connectome analyses in various brain regions (Dolan et al., 2019;Hulse et al., 2021;Ito et al., 2014;Li et al., 2020;Sayre et al., 2021). Future studies on multisensory convergences in newly demarcated multisensory integration centers in the honey bee cerebrum, their feedback connections with sensory processing centers, and their association with premotor centers will be crucial in promoting our understanding of the neuronal mechanisms underlying the stunningly complex behaviors of the honey bee.
In the present study, we provide a comprehensive digital 3D atlas of neuropils in the cerebrum of the honey bee brain. By creating an individual brain atlas, we were able to include many brain neuropils that had not been demarcated in the honey bee average shape atlas (Brandt et al., 2005). Using 3D-reconstructions based on confocal image stacks from whole mount brains immunolabeled with a synaptic marker and comparison with neuronal tracing of major sensory tract systems, we show that almost all neuropils that had been identified in the cerebrum of Drosophila (Ito et al., 2014) could be found, reconstructed, and mapped in the honey bee cerebrum. The resulting detailed 3D atlas of cerebral neuropils and tract systems with previously unknown demarcations in the honey bee will be an important neuroanatomical tool for future structural and functional analyses at multiple levels, for example, the study of multisensory convergence, plasticity, and neuromodulatory circuits in the brain of this charismatic social insect.

Animals
Honey bees (Apis mellifera carnica) were taken from colonies reared in the apiary of the chair of Zoology II at the University of Würzburg.
Bees had free access to natural food sources. For all neuroanatomical

Immunohistochemistry
Bees were anesthetized on ice and afterward decapitated. The head was fixed in dental wax-coated dishes, and a small window was cut between the compound eyes with a razor blade. The staining procedure has been previously described by Hensgen et al. (2021)  Animals were anesthetized on ice and fixed on a holder using dental wax. A small window was cut above the antennae with a razor blade.
Trachea and gland tissues were removed, and the brain was washed in bee ringer solution (130 mM NaCl, 5 mM KCl, 4 mM MgCl 2 , 5 mM CaCl 2 , 15 mM Hepes, 25 mM Glucose, 160 mM Sucrose-diluted in distilled water and adjusted to pH 7.2 with NaOH and HCl). Subsequently, the fluorescently labeled tracer was injected with a glass capillary into the neuropils of interest (antennal lobe, optic lobes). After injection, the brain was again washed with bee ringer solution and the head capsule resealed to avoid dehydration. Samples were incubated for 3-4 h in humidified dark chambers at room temperature before the brain was dissected in bee ringer. Next, brains were fixated overnight in 4% formaldehyde at room temperature and washed 3 × 10 min in PBS. Then, samples were pre-treated with PBS containing 2% Triton X-100 (2% PBST, 1 × 10 min) and or in 0.5% PBST (2 × 10 min) before they were pre-incubated in 0.5% PBST with 2% NGS for 1 h at room

2.5
Laser scanning confocal microscopy, image processing, and data analysis A Leica TCS SP8 confocal laser scanning microscope (Leica Microsystems AG) was used to image the brain samples. For all scans, we used a 20 × water immersion objective (20.0 × 0.7/0.75 NA) with a step size of 5 µm in z-direction and a resolution of 0.75 µm in x-and ydirections. Fluorophores were excited in different combinations with a wavelength of 488 nm for Alexa Fluor 488 dextran, 568 or 633 nm for anti-synapsin, and 568 nm for Microruby. Images were processed using ImageJ (ImageJ 1.52p; Wayne Rasband) and, if necessary, the contrast was enhanced. We used CorelDRAW X8 (Version 20.0.0.633, Corel Corporation, Ottawa, Canada) to arrange all confocal images and 3D reconstructions of the brain.

3D labeling and reconstruction
All 3D reconstructions of brain neuropils and fiber tracts were based on anti-synapsin labeling of an individual brain to obtain an atlas with as many details as possible. Anti-synapsin labeling stains synapserich regions and fiber tracts appear as dark profiles in the brain amira-avizo). The region of interest was manually labeled in the segmentation editor of the Amira software before the 3D models were generated using the SurfaceGen module.

Axes and naming conventions
The nomenclature for the neuronal structures in the bee brain was based on the unified nomenclature for the insect brain (Ito et al., 2014).
We used known fiber bundles as unambiguous landmarks to demarcate individual brain neuropils of the cerebrum. Like in other studies (e.g., Habenstein et al., 2020;Heinze & Reppert, 2012;Immonen et al., 2017), we additionally used structural characteristics and alterations of anti-synapsin labeling in confocal images to discriminate different brain neuropils. In some cases, when no apparent distinction between neuropils was possible, we defined traceable and reproducible landmark criteria to demarcate individual neuropils, which in some cases resulted in relatively straight boundaries between these neuropils. This is consistent with the approach in similar publications about individual brain atlases in different insect species (e.g., Immonen et al. 2017;von Hadeln et al. 2018;Habenstein et al. 2020). We based our definition of the term cerebrum as used in Ito et al. (2014). The cerebrum includes the protocerebrum without the optic lobes, the deutocerebrum, and the tritocerebrum.
For spatial orientation in the honey bee brain, it is important to keep in mind that in Hymenoptera the central brain does not undergo a 90 • rotation during metamorphic development, as it was shown for many other insect species including dipterans, lepidopterans, and coleopterans Heinze & Reppert, 2012;Huetteroth et al., 2010;Immonen et al., 2017;Ito et al., 2014). Due to this 90 • change in brain orientation, the locations of respective brain neuropils are changed accordingly. In addition to the absence of the 90 • rotation in Hymenoptera, the massive enlargement of the mushroom body calyces in the honey bee represents another factor leading to the compression of the superior neuropils underneath the mushroom body calyces. As a result, the actual localization of some cerebral neuropils and their size and shape may differ from their spatial orientation suggested by their nomenclature. For instance, due to this rotation, the superior neuropils are actually located in the anterior region of the honey bee brain.
We do not claim that the respective neuropils are homologous, but we tried our best to reconstruct and assign the names based on close comparison with the Drosophila and other fully reconstructed insect brain atlases. In this study, we use the terms dorsal and ventral with reference to the body axis, rather than the neuroaxis.

General layout of the Apis mellifera brain
The insect brain can be divided into three regions: the protocerebrum, the deutocerebrum, and the tritocerebrum. As in other hymenopteran species (e.g., Brandt et al., 2005;Couto et al., 2016;Groothuis et al., 2019;Habenstein et al., 2020), these three parts are fused into one contiguous cerebral ganglion (CRG) in the honey bee. Moreover, the transition between the CRG and the ventrally attached gnathal ganglion (GNG) is seamless. Some neuropils of the CRG have received greater attention due to well-defined boundaries with adjacent brain areas. These well-defined brain neuropils are also called "major neuropils." The major neuropils include important sensory input areas, but also higher processing and integration centers of the brain. In the honey bee brain, the antennal lobes (AL), the optic lobes (OL), the mushroom bodies (MB), and the central complex (CX) form the major neuropils.
These neuropils have been subject to numerous anatomical and physiological studies, and main components of them have been reconstructed in a 3D average-shape brain atlas (Brandt et al., 2005). In the following, we will give only a brief description of the major neuropils to then focus on the newly demarcated cerebral neuropils.
The lateralmost neuropils on each side of the honey bee brain are the OLs (Figure 2). Each OL receives primary visual input and has a relatively large volume in visually navigating honey bees. The OL is divided into the lamina, medulla, and lobula. Visual information is transferred from the retina to the lamina and the medulla (Ribi, 1975) and further processed in the three optic neuropils. From there, information is finally transmitted via several visual tracts to higher integration centers of the brain (reviewed in Hempel de Ibarra et al., 2014;Hertel & Maronde, 1987). Like the OL, the AL is a very prominent neuropil in the honey bee brain. The AL lies in the ventralmost part of the CRG (Figure 2a,b,d,e) and is separated from the neighboring neuropils by a thick boundary, most likely from neuronal and glial cell bodies and glial processes. The AL is the primary olfactory input region in the honey bee brain. From the AL, olfactory information is transmitted via the antennal-lobe tracts (ALT) to the mushroom bodies and other olfactory processing centers in the cerebrum, most prominently in the lateral horn (Galizia & Rössler, 2010;Kirschner et al., 2006;.
The MB and CX are important high-order integration centers of the honey bee brain. Each of the paired MBs consists of a medial and a lat-

Central adjoining neuropils
The major neuropils are surrounded by many further neuropils with less distinct boundaries. These neuropils are termed central adjoining neuropils (CANP). Since they are not as easy to identify, these neuropils have received much less attention than the major neuropils in the past. However, the CANPs occupy a large volume in the brain of the honey bee and other insects and are thought to be important in multisensory processing and the generation of premotor behavioral output (Habenstein et al., 2020;Heinze & Reppert, 2012;Ito et al., 2014). To be able to better discriminate individual neuropils F I G U R E 2 General layout of the Apis mellifera brain. (a-d) Surface reconstructions of the major neuropils in the honey bee brain. The major neuropils are the antennal lobes (AL), the mushroom bodies (MB), the optic lobes (OL), and the central complex (CX). The three-dimensional reconstruction of the brain neuropils is visualized from an anterior (  within the CANP, we reconstructed prominent fiber bundles and used them as additional landmarks (Figure 3a-e). This procedure has previously been implemented successfully in other anatomical studies on different insect species (Habenstein et al., 2020;von Hadeln et al., 2018;Heinze & Reppert, 2012;Immonen et al., 2017). In anti-synapsin labeled brains, neuronal tracts can be identified as anti-synapsin-free zones. For improved identification of the olfactory and visual tracts, we additionally used anterograde staining of the projection neuron tracts from the AL and OL, respectively ( Figure 1). Together, this allowed the reconstruction of a comprehensive, 3D fiber bundle system. Overall, we reconstructed 25 fiber bundles in the honey bee brain. Using these as landmarks within the anti-synapsin labeling patterns, we identified and localized 16 paired and five unpaired CANPs in the bee brain (see serial optical sections shown in Figure 4, and 3D reconstructions in

Anterior optic tubercle
The anterior optic tubercle (AOTU) receives visual information (chromatic cues and polarized light) from the optic lobes via the anterior optic tract (AOT) (see also Mota et al., 2013Mota et al., , 2011Zeller et al., 2015). are therefore easy to recognize in the honey bee brain.

Posterior optic tubercle
The

Lateral complex
The lateral complexes (LX) are bilateral neuropils that house the bulbs (BU) and the lateral accessory lobe (LAL) (Figure 4d

Superior neuropils
The a connection between the two brain hemispheres in the anteriormost part of the brain. More posteriorly, the median bundle (MBDL) and the ATL separate the SMP at the midline of the brain (Figures 3 and 6b). and ATL (Figures 3, 4e, and 6b).

Inferior neuropils
The antler (ATL), the clamp (CL), and the inferior bridge (IB) form the inferior neuropils (INP) in the honey bee brain (Figure 4d-j). Unlike previously described in the fruit fly Drosophila melanogaster (Ito et al., 2014) or Cataglyphis nodus ants (Habenstein et al., 2020), the crepine was not found as a distinct neuropil in the honey bee brain.

Lateral horn
The lateral horn (LH) is an important high order olfactory processing center in the honey bee brain and receives olfactory input from the ALTs (see also : Abel et al. 2001;Kirschner et al., 2006;.
We therefore defined the outlines of the neuropil based on its innervation by projection neurons of the ALTs. In addition to this, the complete

Ventrolateral neuropils
The are no clear cellular boundaries to these neuropils, we defined the borders based on several fiber bundles or slight differences in the antisynapsin-IR. The PLP forms the dorsal region of the VLNP, whereas the VLP is located more ventrally in the bee brain (Figures 4d-j and 5b-d).
The two neuropils are separated by the ml-ALT 2, the posterior lateral fascicle (PLF, anterior brain region), and the IOC (posterior brain region)

Ventromedial neuropils
The ventromedial neuropils (VMNP) consist of the posterior slope (PS) and the ventral complex (VX) (Figure 4g-l). In Drosophila, the VX is further subdivided into the vest, the gorget, and the epaulette, while the PS has a separate superior (SPS) and inferior (IPS) subregion (Ito et al., 2014). In honey bees, however, these brain regions appear rather contiguous, and VX and PS are not separable into subneuropils without further labeling tools. This is in line with the findings in the ant C. nodus (Habenstein et al., 2020). In the honey bee, the VMNPs flank the posterior part of the esophagus (Figure 4g-l). The anterior edge of the VMNPs lies immediately behind the LAL and extends over large parts of the bee brain to the GNG. Within the VMNP, the VX is the more anteriorly located brain neuropil. Anterior and posterior ends of the VX coincide with the appearance of the IFS, which additionally demarcates the lateral border to the VLP together with the l-ALT (Figure 7b).
Therefore, the IFS is the most important landmark for localizing the VX in the honey bee brain. Due to the dorsal demarcation by the IOC from the adjacent CL and IB, the VX occupies a well-defined space between the esophagus and l-ALT. Only the ventral boundary to the SAD appears rather fluid and was therefore determined based on slight differences in anti-synapsin-IR (Figures 4g-h and 5c). The PS has a much larger volume compared to the VX and is located posteriorly to the VX.
The PS flanks the esophagus from its superior to its inferior border (Figure 4i-l). Therefore, the esophagus serves as the main landmark to set the borders to dorsally (IB, CL) and ventrally (SAD, GNG) neighboring neuropils. In addition, the MEF serves as a dorsal border to the CL.
In the anterior region of the PS, the CTT provides a lateral demarcation to the VLNPs (Figure 7b). More posteriorly, once the VLNPs and the LEF disappear, the PS occupies the entire region between the esophagus and lateral edge of the cerebrum (Figures 4k and 7b2). Similar to findings in Drosophila (Ito et al., 2014) and Cataglyphis ants (Habenstein et al., 2020), the posterior optic commissure (POC) runs through the region of the PS in the honey bee brain (Figure 7b).

Periesophageal neuropils
The periesophageal neuropils (PENP) comprise four distinct neuropils: the cantle (CAN), the flange (FLA), the prow (PRW), and the saddle (SAD) (Figures 4d-i and 5a,b,d). These brain neuropils are in the ventralmost region of the cerebrum. FLA and CAN are both relatively small neuropiles and flank the dorsal part of the (anterior) esophagus. As in the brain of Cataglyphis ants (Habenstein et al., 2020), the FLA is a very thin and elongated neuropil in the honey bee brain. It extends from the root of the MBDL to the dorsomedial tip of SAD and PRW (Figure 7c), which is in line with findings in other insect species (Habenstein et al., 2020;Ito et al., 2014). The lateral and medial borders of the FLA are clearly demarcated by cellular boundaries, most likely glial processes.
While medial to the FLA only the esophagus is localized, there are three laterally adjacent neuropils: the SMP (anterior), the LAL, and the SAD (posterior). The cellular architecture and function of the SAD in the honey bee brain are not well known, although they are likely to also contain projections from antennal mechanoreceptors as described in the honeybee (Ai et al. 2007) and in Catagylphis ants (Grob et al. 2021;Habenstein et al. 2020). The ventral transitions to SAD and PRW are a bit more difficult to identify in the honey bee brain. Here, the transitions occur relatively seamlessly at the level of the superior edge of the antennal mechanosensory and motor center (AMMC) (Figure 7c1).
The CAN is a small, triangular-shaped neuropil with clearly visible borders to the neighboring neuropils. It is located ventral to the MB medial lobes, medial to the LAL, dorsal to the posterior tip of the FLA, and anterior to the IB. In contrast to FLA and CAN, SAD and PRW occupy large areas in the bee brain. The SAD further includes a prominent subneuropil-the AMMC (Figures 4c,d and 7c1). This neuropil plays a central role in insects as it represents an important sensory input region. The AMMC receives primary gustatory and mechanosensory input from the antennae (Ehmer & Gronenberg, 2002;Farris, 2008;Grob et al., 2021;Haupt, 2007;Kamikouchi et al., 2009;Miyazaki & Ito, 2010). In honey bees, the boundaries of the AMMC was determined via anterograde fluorescent tracing of the antennal nerves (see Ai et al., 2007;Ai & Hagio, 2013;Kumaraswamy et al., 2019). Here, the anterior tip of the AMMC is in the center of the AL (Figure 4c,d). More posteriorly, the transition is smooth to the remaining part of the SAD, which encloses the AMMC laterally, ventrally, and medially. Dorsally, m-ALT and l-ALT demarcate the AMMC from the adjacent LAL. The AMMC of both brain hemispheres are connected via the AMMCC (Figure 7c). The name of the SAD is derived from its shape and location in insect brains.
Here, the SAD lies like a saddle on dorsal part of the GNG (Ito et al., 2014). In honey bees, the anteriormost part of the SAD is found in the to the AL, the AMMC, and the SAD. In this part of the brain, the PRW has a triangular shape and is ensheathed by thick processes from most likely glial cells. In its posteriormost region, the neuropil is connected across the midline of the brain. In this region, the junctions with the neighboring SAD (dorsolateral) and GNG (lateral, posterior) are fluid.

General aspects of a 3D atlas of the honey bee central brain
We reconstructed 35 neuropils in the honey bee brain together with 25 major neuronal fiber tracts and commissures that run through the cerebrum. This adds substantial detail to the information provided by the average shape atlas from Brandt et al. (2005). The average shape brain F I G U R E 7 Three-dimensional reconstruction of individual central adjoining neuropil (CANP) groups and associated fiber bundles (continuing from Figure 6). All CANP neuropil groups are visualized in anterior (1), posterior (2), and anterolateral view (3). (a) Ventrolateral neuropils (VLNP), lateral horn (LH), and associated fiber bundles. The VLNP consist of the posterolateral protocerebrum and the ventrolateral protocerebrum. Important fiber bundles for the demarcation of these neuropils are the clamp-tritocerebral-tract (CTT), the horizontal ventrolateral protocerebrum fascicle (hVLPF), the inferior fiber system (IFS), the inferior optic commissure (IOC), the lateral antennal lobe tract (l-ALT), the lateral equatorial fascicle (LEF), the mediolateral antenna lobe tract 2 (ml-ALT2), the posterior lateral fascicle (PLF), and the superior posterolateral protocerebrum commissure (sPLPC). (b) Ventromedial neuropils (VMNP) and associated fiber bundles. The posterior slope (PS) and the ventral complex (VX) form the VMNPs. The CTT, the IFS, the IOC, the l-ALT, the medial equatorial fascicle (MEF), and the posterior optic commissure (POC) (Continues) F I G U R E 7 (Continued) were used to define the borders of the VMNP. (c) Periesophageal neuropils (PENPs) and associated fiber bundles. The PENPs comprise the antennal mechanosensory and motor center (AMMC), the cantle (CAN), the flange (FLA), the prow (PRW), and the saddle (SAD). For the demarcation of the PENPs, the antennal mechanosensory and motor center commissure (AMMCC), the medial antennal lobe tract (m-ALT), the median bundle (MBDL), and the l-ALT were used. Scale bars = 200 µm.
atlas by Brandt et al. (2005) was restricted to major neuropils such as the AL, CB, MED, LOB, and individual compartments within the medial and lateral MB. Similarly, a recent average shape brain atlas of the bumblebee, based on micro computed tomography data, was restricted to these major neuropils (Rother et al., 2021). Recent studies suggest that so far neglected neuropils in the cerebrum serve important functions in multisensory integration (Currier & Nagel, 2020). For example, whole brain imaging with novel microscopic techniques such as light field microscopy in behaving insects are beginning to elucidate the function of so far unknown neuropils in the central brain (e.g., Aimon et al., 2019Aimon et al., , 2022. However, the specific function and circuit architectures of cerebral neuropils, in many cases, still remain "terra incognita" (Ito et al., 2014).
Our present 3D atlas of neuropils in the central brain of the honey bee is intended to provide an anatomical platform that can be used to map the projection patterns of neurons with known origin and function (e.g., sensory circuits) to localize the distribution of neurotransmitters and neuromodulators (e.g., , anatomically map data from functional imaging studies (e.g., live calcium imaging), or identify brain regions that undergo behavioral stage-, age-or experience-dependent structural plasticity (Groh & Rössler, 2020). Further potential applications include mapping the arborization of individual neurons obtained from combined electrophysiological recordings and neuronal tracing experiments, or from activity-driven expression of immediate early genes and their gene products (Sommerlandt et al., 2019(Sommerlandt et al., , 2017. So far, many studies were mostly restricted to the major neuropils. While an average-shape atlas is very useful to register single neurons from other individuals into an average brain, the individual brain atlas provides a more comprehensive atlas with more detail and higher spatial resolution of 3D reconstructions of even small neuropils and their boundaries. As many neuropils in the honey bee cerebrum have small volumes and complex, thin, or convoluted structures, an average shape atlas for these neuropils is not feasible because shapes and boundaries of such neuropils become distorted or blurred in an average-shape 3D reconstruction, especially when obtained at a low optical resolution. The same is true for the reconstructions of relatively thin tracts with complex shapes. However, fine-grained neuropil and tract atlases are needed for studies of circuits and their connectivity and function. For this reason, recent studies increasingly pursued 3D atlases from individual brains of various insect species comprising comprehensive 3D neuropil maps and connecting fiber tracts at high spatial resolution (e.g., Habenstein et al., 2020;von Hadeln et al., 2018;Heinze & Reppert, 2012;Immonen et al., 2017;Ito et al., 2014;Kaiser et al. 2022; for further examples, see the Insect Brain Database website https://www.insectbraindb.org/app/). Individual brain atlases, however, come with the disadvantage that they represent only one specimen. A similar problem had to be faced in recent connectome approaches based on serial electron microscopy data from single specimen (e.g., Hulse et al., 2021;Ito et al., 2014;Scheffer et al., 2020). This limitation may be overcome by future developments in high-resolution imaging techniques and high-throughput histological protocols. This may allow the comparison of multiple anatomical datasets to achieve a better estimation of inter-individual variation and its role in neuronal computation and behavior.
The main strategy for our individual 3D brain atlas of the brain of a honey bee worker was to identify transparent and clear definitions of 3D neuropil boundaries and their spatial relationships with adjacent neuronal fiber tracts. Some challenges and limitations of this method have been outlined in Section 2.6 and may hopefully be overcome in the future by novel neuropil specific molecular markers and other tracing tools. The resulting comprehensive 3D reconstructions of neuropil maps and fiber tracts can be easily used to identify the projection fields of individually traced neurons or bulk-loaded neuronal populations (mass fills) using, for example, side-by-side comparison of confocal image stacks from tracings that have been combined with anti-synapsin labeling. This is possible with image stacks from a whole mount brain and using sectioned material. With such an approach, it was recently possible to precisely determine the convergences of mechanosensory afferents from the Johnston's organ with projections from ocellar neurons and visual input from the optic lobes in the posterior neuropils (PS, VX, and VLP) within the cerebrum of the desert ant brain (Grob et al., 2021;Habenstein et al., 2020). A similar strategy was recently used to identify the specific target neuropils and projection patterns of neuropeptidergic neurons in the central brain of Cataglpyhis nodus (Habenstein, Schmitt, et al., 2021;Habenstein, Thamm, et al., 2021).
The insect brain database (https://www.insectbraindb.org) contains both average-shape and individual brain atlases and, thus, provides a unique platform for 3D reconstructed insect brains . Consequently, our present detailed 3D neuropil atlas of the honey bee cerebrum can easily be used in combination with the average shape atlas by Brandt et al. (2005), the high-resolution averageshape atlas of the CX (Kaiser et al., 2022), and future high-resolution maps of individual brain neuropils, neuronal tracts, or individual neurons that have been registered and added to existing atlases in the honey bee.
Most neuropils and tracts in the honey bee cerebrum have counterparts in the brain of the fly (Ito et al., 2014), dung beetle (Immonen et al., 2017), locust (von Hadeln et al., 2018, and desert ant (Habenstein et al., 2020). However, differences in the spatial arrangement of neuropils in the honey bee brain become especially evident in comparison with dipteran, lepidopteran, and coleopteran species Heinze & Reppert, 2012;Immonen et al., 2017;Ito et al., 2014). This is because hymenopteran brains do not undergo a 90 • rotation during metamorphic development. As a result, the central brain of Hymenoptera, but also of hemimetabolous insects, is oriented perpendicular to the fly brain (e.g., Brandt et al., 2005;Groothuis et al., 2019;Ito et al., 2014;Kurylas et al., 2008). Therefore, the location of some neuropils differs from the ones indicated by their names (e.g., the superior protocerebral neuropils lie in an anterior position in the honey bee). For the sake of comparability between insect species and similar to our previous study in the desert ant, we adhere to the unified nomenclature introduced by Ito et al. (2014). The general layout of the cerebral neuropils in the honey bee shows a high level of similarity with the brain of the desert ant, C. nodus (Grob et al., 2021;Habenstein et al., 2020). However, the relative size and shape of individual neuropils show distinct differences compared to other ant species. For example, the relative size of the optic lobes is small in strongly olfactory-oriented ant species (Bressan et al., 2014;Groh et al., 2014;Ilieş et al., 2015;Yilmaz et al., 2016). Since major primary sensory neuropils (AL, OL) and integration centers (MB, CX) in the honey bee have been described in detail in other studies (for a most recent 3D reconstruction of the CX, see Kaiser et al., 2022; for a recent review on the MB structure and plasticity, see Groh & Rössler, 2020), we focus the following discussion on the newly demarcated neuropils in the cerebrum.

Central adjoining neuropils
Most neuropils found within the central adjoining neuropil (CANP) of Drosophila (Ito et al., 2014) could also be identified in the honey bee brain. Although the CANPs fill a large proportion of the brain in the honey bee and other insect species, their specific function is not well known. Mostly based on anatomical criteria like sensory tracts or the morphology of individual neurons targeting these neuropils, they have been assigned an important role in sensory integration and their connections to centers that are involved in the control of motor output (Habenstein et al., 2020;Heinze & Reppert, 2012;Ito et al., 2014). This is, for example, the case for the AOTU or the bulbs and other parts of the lateral complex, but much less clear for the POTU (Held et al., 2016;Kaiser et al., 2022;Mota et al., 2011;Zeller et al., 2015). For other regions of the CANP, it is much less clear for most insects, but a recent whole brain calcium imaging study in behaving Drosophila, for example, could demonstrate the activation of several neuropils within the CANP during walking behavior (Aimon et al., 2019).
Our anatomical demarcation of individual CANP neuropils in the honey bee brain was only possible by using 25 prominent neuronal fiber bundles as landmarks (identified and 3D reconstructed as antisynapsin free zones or via tracing experiments; Figure 1). Overall, we identified 16 paired and five unpaired CANPs in the honey bee brain

Anterior and posterior optic tubercle
The anterior optic tubercle (AOTU) in the honey bee is divided into an upper and lower unit, which appears to be a common feature that has been found in all insects studied so far Habenstein et al., 2020;von Hadeln et al., 2018;Heinze & Reppert, 2012;Homberg et al., 2004;Immonen et al., 2017;Montgomery & Merrill, 2017;Mota et al., 2011;Pfeiffer & Kinoshita, 2012;Strausfeld et al., 2007). While the lower unit of the AOTU is part of the sky-compass pathway, the upper unit of the AOTU is involved in processing of chromatic light cues (e.g., Mota et al., 2013) and plays a role in figure-ground discrimination in flies (Aptekar et al., 2015). The AOTUs on both sides are connected via the TUBUT to the BU (see below). Interestingly, the relative size of the AOTU appears larger in the honeybee compared to Cataglyphis ants (Habenstein et al., 2020). In the Cataglyphis brain, the posterior optic tubercle (POTU) could not be demarcated as a distinct anatomical structure. In contrast, in the honey bee, it is a distinct structure located in the posteriormost region of the brain and attaches postero-

Lateral complex
The bilateral lateral complexes (LX) comprise the bulbs (BU) and the lateral accessory lobe (LAL). In the honeybee, there are two distinct bulbs that contain conspicuously large microglomerular synaptic structures of variable size (Held et al., 2016;Mota et al., 2016). The LAL is relatively large and has been shown to be a major target of the CX output (e.g., Hensgen et al., 2021;Homberg et al., 2011;Kaiser et al., 2022). In Cataglyphis, the number of bulb microglomerular complexes was shown to increase upon first exposure to light, especially UV light, during a sensitive period of the ants' individual ontogeny at the transition from the dark interior of the nest to outside foraging (Grob et al., 2022;Schmitt et al., 2016). Whether this is also the case for bulb microglomerular complexes in the honey bee and other insect species still needs to be investigated. In contrast to the anatomically conspicuous bulbs, the anatomical details of the LAL in the honey bee brain have hitherto remained undescribed due to its diffuse shape and difficult localization. It is known from other insects like the locust and Drosophila that the LAL is innervated by output neurons from the CX (Hulse et al., 2021;Kaiser et al., 2022;, but also receives input from ascending neurons Homberg, 1994).
The LAL also acts as a premotor center as it is also targeted by descending neurons that relay motor-control information to the thoracic ganglia, which was shown in Bombyx mori and Drosophila melanogaster Namiki & Kanzaki, 2016;.

Superior neuropils and inferior neuropils
The superior neuropils ( are shifted downward to the anterior brain compared, for example, to the situation in the fly brain (Ito et al., 2014). The SMP stretches between both brain hemispheres, whereas the SIP wraps around the vertical lobe (VL) of the MB on both sides. This region has previously been termed "ring neuropil" (e.g., Kirschner et al., 2006). All superior neuropils in the honey bee are populated by arborizations of MB output neurons (Rybak et al., 2010;Rybak & Menzel, 1993;Strausfeld, 2002). Recent connectome studies in Drosophila show that a substantial proportion of MBONs has synaptic contacts with fan-shaped body neurons of the CX, mostly within the SMP (Hulse et al., 2021;Li et al., 2020). In the honey bee, no further information about downstream neurons in the SNP is available. However, we can assume that the SNP represents a major hub for preprocessed sensory input and the information from MB memory circuits and other regions with premotor areas such as the LAL together with posterior and ventral neuropils (e.g., PS, VLP) that are innervated by descending neurons Namiki & Kanzaki, 2016). Interestingly, four large corazonergic neurons on each side of the brain in Cataglypyhis were shown to have prominent arborizations in the SMP of both hemispheres and exhibit characteristic projections into the neurohormonal organs of the retrocerebral complex, which connects this region with hormonal systems .
The inferior neuropils (INP) in the honey bee comprise the antler (ATL), the clamp (CL), and the inferior bridge (IB). Unlike in Cataglpyhis (Habenstein et al., 2020), the crepine was not found as a distinct neuropil in the honey bee. As in dung beetles (Immonen et al., 2017), the ATL is the anteriormost neuropil and extends across the midline. The function of the INPs in the honey bee are not known so far.

Lateral horn
The LH is a prominent target neuropil of output tracts from all types of AL projection neurons along the medial (m-ALT), mediolateral (ml-ALT), and lateral (l-ALT) antennal lobe tracts (Abel et al. 2001;Kirschner et al., 2006;Zwaka et al., 2016). Three ml-ALTs comprise axons from multiglomerular AL projection neurons that innervate multiple glomeruli and project to the middle and ventral parts of the LH. A calcium imaging study in the honey bee has shown that odor-similarity relationships within the AL are largely conserved, whereas information about pheromones remains segregated in the LH (Roussel et al., 2014). In addition to the LH, ml-ALT neurons target further neuropils in various parts of the superior protocerebrum (SIP, SLP), regions that had originally been termed "ring neuropil," "triangle," and "lateral bridge" (Kirschner et al., 2006). Interestingly, the LH, SMP, SIP, and SLP in the honey bee are also targeted by several types of mushroom body output neurons (MBONs) with a majority targeting the SMP (Rybak & Menzel, 1993;Strausfeld, 2002). Olfactory processing in the LH and neighboring neuropils, therefore, is likely to receive recurrent modulatory feedback from MBONs. This may also be true for MBONs responding to visual or multisensory input. In addition to multisensory information, MBONs may even mediate recurrent feedback of memory-associated information (Strube-Bloss et al., 2012;Strube-Bloss & Rössler, 2018) to the LH and the superior neuropils.
Long-range recurrent modulatory feedback by MBONs to upstream olfactory processing regions in the LH and surrounding neuropils targeted by olfactory tracts from the AL represent an interesting topic for future studies on context and experience dependent modulation of olfaction in the honey bee.

Ventrolateral neuropils, ventromedial neuropils, and periesophageal neuropils
Within the periesophageal neuropils, the antennal mechanosensory and motor center (AMMC) represents a prominent neuropil located dorsal to the AL that receives input from antennal mechanoreceptors, most importantly from the Johnston's organ (JO). Sensory afferents from receptor neurons of the JO in Cataglyphis ants have recently been mapped to the AMMC, the saddle (SAD), ventrolateral protocerebrum (VLP), and the posterior slope (PS) (Grob et al., 2021). Similar projections can be deduced from previous work in the honey bee, although these were only based on 3D mapping into the atlas by Brandt et al. (2005) and the ventromedial and ventrolateral neuropils can only be estimated from these studies, as these cerebral neuropils had not yet been demarcated at the time (Ai et al., 2007(Ai et al., , 2017. Interestingly, in both Catagylphis and the honey bee mechanosensory afferents from the Johnston's organ converge with visual interneurons from the ocelli and from the optic lobes in the PS, ventral complex (VX), and VLP (Ai et al., 2007;Grob et al., 2021;Habenstein et al., 2020). This suggests that these neuropils can be regarded as multisensory integration centers (Currier & Nagel, 2020). Furthermore, they can also be seen as premotor centers, as, together with the LAL, the PS and VLP were shown to be innervated by descending neurons in both the silk moth and Drosophila (Namiki, Dickinson, et al., 2018;Namiki & Kanzaki, 2016;. Finally, connections of both neuropil regions to the CX seem likely from studies on auditory and wind compass circuits arising from the Johnston's organ (Lai et al., 2012;Okubo et al., 2020). The functions of most of the other periesophageal neuropils like CAN, FLA, PRW, and the PLP are largely unknown in the honey bee and other insects.
The availability of new fast live-imaging techniques such as light-field microscopy (Aimon et al., 2019(Aimon et al., , 2022 will eventually allow imaging of the entire brain in a behaving honey bee. Combinations with highresolution tracing and connectome studies of circuits at the ultrastructural level will promote our understanding of the complex structures in the central brain of this social insect. The present 3D atlas of the honey bee cerebrum represents an important step toward this goal.

ACKNOWLEDGMENTS
We thank Dirk Ahrens for beekeeping and providing honey bees.
Special thanks go to Erich Buchner and Christian Wegener for kindly providing the anti-synapsin antibody. We are grateful to Katja Tschirner for conversion and upload of data to the insect brain data base, and we especially thank Ronja Hensgen for her invaluable help with the the insect brain database website. We also thank two anonymous reviewers for their helpful suggestions to improve this manuscript and implementation of 3D data into insect brain Open access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest

DATA AVAILABILITY STATEMENT
Three-dimensional data of the brain of the honey bee are available at the Insect Brain Database website (https://www.insectbraindb.org/).

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/cne.25486.