Comprehensive map of visual projection neurons for processing ultraviolet information in the Drosophila brain

Abstract The brain perceives visual information and controls behavior depending on its underlying neural circuits. How UV information is represented and processed in the brain remains poorly understood. In Drosophila melanogaster, UV light is detected by the R7 photoreceptor that projects exclusively into the medulla layer 6 (M6). Herein, we imaged 28,768 single neurons and identified 238 visual projection neurons linking M6 to the central brain. Based on morphology and connectivity, these visual projection neurons were systematically classified into 94 cell types belonging to 12 families. Three tracts connected M6 in each optic lobe to the central brain: One dorsal tract linking to the ipsilateral lateral anterior optic tubercle (L‐AOTU) and two medial tracts linking to the ipsilateral ventral medial protocerebrum (VMP) and the contralateral VMP. The M6 information was primarily represented in the L‐AOTU. Each L‐AOTU consisted of four columns that each contained three glomeruli. Each L‐AOTU glomerulus received inputs from M6 subdomains and gave outputs to a glomerulus within the ellipsoid body dendritic region, suggesting specific processing of spatial information through the dorsal pathway. Furthermore, the middle columns of the L‐AOTUs of both hemispheres were connected via the intertubercle tract, suggesting information integration between the two eyes. In contrast, an ascending neuron linked each VMP to all glomeruli in the bulb and the L‐AOTU, bilaterally, suggesting general processing of information through the ventral pathway. Altogether, these diverse morphologies of the visual projection neurons suggested multi‐dimensional processing of UV information through parallel and bilateral circuits in the Drosophila brain.

(EB), a ring-shaped structure located at the center of the brain, acts as an internal compass for navigation. The neural activity in EB neurons reflects head orientation during locomotion (Fisher et al., 2019;Kim et al., 2019). However, how UV information is represented in the brain and relayed to the EB remains poorly understood.
The axons of UV-sensitive R7 neurons terminate exclusively within the layer 6 of the medulla (M 6 ). Serial-section electron microscopy revealed that amacrine Dm8 neurons with dendritic arbors in the M 6 receive direct synaptic inputs from R7 neurons and connect with transmedulla Tm5c neurons at the lobula. Consistently, functional studies showed that information flows via R7 ! Dm8 ! Tm5c are necessary for the UV preference behavior in Drosophila (Gao et al., 2008;Takemura et al., 2013;Karuppudurai et al., 2014). Conversely, anatomical and functional studies have revealed that other M 6 output neurons known as MT (also, MC61 or MeTu) terminate at the lateral anterior optical tubercle (L-AOTU) are necessary for the phototaxis toward UV light (Otsuna et al., 2014). These results suggested that UV information is diverged from M 6 by different downstream circuits related to control-specific aspects of behavior.
However, the M 6 output circuits for processing UV information in the Drosophila brain remain to be fully determined.
In this study, we performed a large-scale 3D imaging of single neurons connecting the UV-receiving M 6 to the central brain in Drosophila. Our results on the topographical organization of 238 reconstructed M 6 downstream neurons suggested that UV inputs are differentially and hierarchically represented in the brain. Furthermore, the UV information received by the two eyes is integrated and processed in a multilayer manner.

| Fly stocks
Fly strains (Drosophila melanogaster) were reared on a cornmeal-yeast-agar medium at 25 C with 70% relative humidity under a 12 h/12 h light/dark cycle. Wild type Canton-S flies were used for the generation of the representative model of anterior optic tubercle. The GR-Gal4 and GR-LexA lines were ordered from the Bloomington Drosophila Stock Center and the VT-Gal4 lines from the Vienna Drosophila Resource Center.

| Immunohistochemistry
Brain samples were dissected within cold isotonic phosphate-buffered saline (PBS) and fixed immediately in 4% paraformaldehyde in PBS on ice with microwave irradiation (700 W, 90 s, three times). The samples were fixed again in 4% paraformaldehyde in PBS with 0.25% Triton X-100 for another session of microwave irradiation (700 W, 90 s, three times). Moreover, the fixed samples were degassed by vacuum in PBS containing 2% Triton X-100 and 10% normal goat serum (NGS) for 1 h to expel air from the tracheal system and then blocked; then the brain samples were penetrated in the PBS containing 2% Triton X-100 and 10% NGS at 4 C overnight. For immunohistochemistry, brain samples were incubated in PBS containing 0.25% Triton X-100 and 1% NGS with mouse anti-discs large (DLG) antibodies (1:50, antibody 4F3; Developmental Studies Hybridoma Bank [DSHB], University of Iowa, IW) for 2 days at 4 C; biotinylated goat anti-mouse IgG antibodies (1:250, Molecular Probes, Eugene, OR) overnight at room temperature; and Alexa Fluor 635 streptavidin (Molecular Probes, Eugene, OR) overnight at room temperature. Brain samples were extensively washed in PBS containing 1% Triton X-100 and 3% sodium chloride at room temperature for 20 min three times between each step.
Finally, the immunolabeled brains were cleared and mounted in the FocusClear™ (CelExplorer, Taiwan) for confocal imaging.

| RESULTS
3.1 | Neural tracts that convey M 6 visual information to the central brain UV-sensing R7 neurons send axonal terminals to M 6 . Through the analysis of spatial connectivity between the neuropils of 28,608 single neurons collected in the FlyCircuit database, we constructed a comprehensive map of the downstream circuitry of R7 neurons ( Figure 1a). The downstream neurons of the M R6 were classified into six families, based on their neuropil connectivity differences . Three neuronal families diverged from the M 6 : one family terminated in the ipsilateral L-AOTU via the dorsal tract (named the MT tract, Figure 1b); another family terminated in the ipsilateral VMP via the short medial tract (named the short MV tract, Figure 1c); finally, another family terminated in the contralateral VMP via the long medial tract (named the long MV tract, Figure 1d). Two other neuronal families diverged from the L-AOTU: one terminated in the ipsilateral BU (named the TB tract, Figure 1e) while the other in the contralateral L-AOTU (named the TT tract, Figure 1f). The remaining neuronal family branched from the VMP and bilaterally terminated to the BU and the L-AOTU of both brain hemispheres (named the VBT tract, Figure 1g). Thus, UV information at M R6 was relayed to three pairs of neuropils (i.e., L-AOTU RL , VMP RL , and BU RL ) by six neuronal families. Meanwhile, M L6 downstream neurons exhibited mirror circuitry of M R6 downstream neurons. Altogether, these results suggested that each of these six neuropils receives UV information from both eyes.
Upon examining the expression patterns of more than 5000 Gal4 and LexA drivers, we identified 22 with a specific expression in MT, TB, or TT neurons ( Figure 2 and Table 1). No driver showed specific expression in short MV, long MV, or VBT neurons. To test neuronal connectivity between R7 neurons, MT, TB, and TT neurons, we used the two-color labeling with selected Gal4 drivers expressed in red fluorescent proteins (mKO) and LexA drivers expressed in green fluorescent proteins (GFP). Two-color labeling demonstrated the physical connections that M 6 input R7 neurons overlapped with M 6 output MT neurons in the M 6 ( Figure 3a1), L-AOTU input MT neurons overlapped with L-AOTU output TB neurons in the L-AOTU (Figure 3b1), and L-AOTU commissural TT neurons overlapped with L-AOTU output TB neurons in the two L-AOTUs (Figure 3c1, d1). To further confirm synaptic connectivity between pre-and post-synaptic neurons, we used the method of green fluorescent protein reconstitution across synaptic partners (GRASP) with selected Gal4 and LexA drivers respectively expressed two split-GFP fragments-spGFP 1-10 and spGFP 11 (Feinberg et al., 2008). GRASP signal was observed in the these findings suggested that UV information is conveyed from the medulla of the optic lobe to the EB of the central brain via two UV F I G U R E 1 Neural tracts that convey medulla layer 6 (M 6 ) visual information to the central brain.

| The L-AOTU consists of four columns
The L-AOTU is located at the dorsal-frontal surface, between the mushroom body lobes and the lateral horn in each hemisphere ( Figure 4a). Previous studies defined three columns in L-AOTU: the medial L-AOTU, the intermediate lateral L-AOTU, and the lateral L-AOTU (Omoto et al., 2017;Timaeus et al., 2020). Here, we performed immunolabeling with anti-DLG antibodies using high-resolution confocal microscopy revealed that the L-AOTU is subdivided into four vertical columns by the aggregation of DLG synaptic proteins (Figure 4b).

| Cell typing and nomenclature
We imaged the 160 single neurons that were stochastically labeled by MARCM (Lee & Luo, 1999) using specific Gal4 drivers expressed in  (Table 3) into 94 cell types (including predicted types), belonging to 42 subfamilies, 12 families, 6 classes, and 3 superclasses ( Figure 6 and Table 3). Neurons of the same cell type were morphologically indistinguishable, connected the same neuropils, and had a similar spatial distribution for both dendritic and axonal F I G U R E 2 Expression patterns of Gal4 and LexA drivers in MT, TB, and TT neurons. (a1-a10) Expression of medulla layer 6 (M 6 ) output MT neuron drivers. (b1-b10) Expression of lateral anterior optic tubercle (L-AOTU) output TB neuron drivers. (c1-c2) Expression of L-AOTU output TT neurons drivers. GFP expression patterns (green) and names of Gal4 and LexA drivers (bottom). Brain neuropils were immunostained with anti-DLG antibodies (blue). Images are frontal views of confocal projections of several adjacent optical sections unless otherwise specified. Scale bar: 20 μm [Color figure can be viewed at wileyonlinelibrary.com] terminals. Neurons with dendrites and axons that innervated the same neuropils belonged to the same family. Within the same family, neurons with axonal terminals distributed in the same neuropil subdivision belonged to the same subfamily. Neurons with dendrites that innervated the same neuropil belonged to the same class. Neurons with mirror morphology between the two hemispheres belonged to the same superclass.
For the systematic analysis of 238 single neurons, each cell type was assigned a simple name based on their dendritic and axonal distri- Table 4, and Table 5) represents the MT neuron with dendritic arbors in the dorsal region of the right M 6 and the axon terminals in the dorsal glomerulus of the right L-AOTU 2 . Upon examining thousands of single-neuron images and driver expressions, we found that these visual projection neurons (VPNs) were highly stereotyped and leftright symmetric in all cases from different individual neurons. This allowed us to predict additional 34 types of M 6 downstream neurons from the specific MT 4 drivers, while and M L6-d,v ! VMP L-s,i was predicted from the symmetrical neuronal partner, M R6-d,v ! VMP R-s,i ).
Using this systematic nomenclature system, we provided a detailed list of all identified and predicted cell types of the M 6 downstream neurons (See Tables 3 and 4 for VPN nomenclature and synonyms).

| MT neurons
All MT neurons linked the M 6 to the ipsilateral L-AOTU and shared similar morphology (Figure 7a

| TT neurons
All TT neurons connected two L-AOTUs between the two brain hemispheres ( Figure 8n). Each TT neuron had dendritic arbors in the multiple glomeruli of two L-AOTU middle columns and axon terminals in the same glomeruli of the contralateral L-AOTU. Based on the differences in the connection between the L-AOTU glomeruli of the two hemispheres, we classified TT neurons into two cell types belonging to one family in each hemisphere ( Figure 6 and Tables 3 and 4). One

| VBT neurons
The identification of the MV neurons allowed us to ask how visual information is conveyed to the EB from the VMP. Through the analysis of thousands of single neurons, we found an ascending neuron, the unique VMP output neuron, linking the unilateral VMP to the bilateral BU and the bilateral L-AOTU (named VBT, Figure 10a). The

| DISCUSSION
In this study, we provided a comprehensive map of single VPNs of the M 6 downstream circuitry in the Drosophila brain. We identified and classified 238 VPNs into 94 cell types belonging to 12 families, based on their 3D morphology. Classification of these single neurons based on connectivity and specific Gal4 expression allowed us to refine distinct partitions of brain regions involved in UV information processing.
The stereotyped circuit organization suggested that environment UV light is hierarchically represented in the brain, which integrates and F I G U R E 5 Double labeling revealed the lateral anterior optic tubercle (L-AOTU) contains four columns. Double labeling of column-specific MT and TB drivers. (a) The axonal innervation of MT 2 is in L-AOTU 2 (labeled with GR52837-LexA, green) and the axonal innervation of MT 2,3 is in L-AOTU 2,3 (labeled with R56F07-Gal4, magenta), confirming previous findings that the intermediate lateral L-AOTU is L-AOTU 2 and L-AOTU 3 . (b) The axonal innervation of MT 2 is in L-AOTU 2 (labeled with GR52837-LexA, green) and the axonal innervation of MT 4 is in L-AOTU 4 (labeled with R73C04-Gal4, magenta), confirming previous findings that the lateral L-AOTU is L-AOTU 4 . (c) The axonal innervation of MT 2 is in L-AOTU 2 (labeled with GR52837-LexA, green) and the dendritic innervation of T 1 B is in L-AOTU 1 (labeled with VT8684-Gal4, magenta).
T A B L E 3 Neurons are identified from Flycircuit (orange), selected Gal4 (blue), and predicted type (gray) [Color T A B L E 3 Continued processes UV information in a multilayer, differential, and bilateral manner.

| Homologous VPNs in other insect species
We morphologically identified six groups of visual projection neurons (VPNs) that connect the optic lobe to the central brain in Drosophila (Figure 1b-g). Several VPNs presented in this study have been previously described in various insect species, including locust (Homberg et al., 2003), butterfly (Heinze et al., 2013), bee (Pfeiffer & Kinoshita, 2012;Zeller et al., 2015), and silk moth (Namiki & Kanzaki, 2018). After performing comparative analysis of the neuronal morphology in the identified M 6 downstream neurons, we provided a detailed list of the homologous VPNs in other insect species (Table 6). Three groups of VPNs are highly similar to the well-characterized polarizationsensitive neurons in locust (Homberg et al., 2003), butterfly (Heinze et al., 2013), and bees (

| Internal representation of UV information at the L-AOTU
Three independent lines of evidence showed that each L-AOTU consisted of four columns, and every column consisted of three glomeruli: field. However, only the two middle L-AOTU columns also receive inputs from the contralateral visual field, suggesting a differential role of the four L-AOTU columns in the integration of UV information from the two eyes.
Furthermore, we predicted additional 34 types of M 6 downstream neurons by examining thousands of single-neuron images and driver expressions (See Table 4 for the list of predicted cell types in each family). These predicted neurons can be grouped into two groups comprised 32 and two cell types that were predicted from the specific drivers and the symmetrical neuronal partner, respectively. However, it should be noted that the neuronal types in each family might be incomplete, although we described a large number of visual projection neurons in this study. Neuronal types that connect L-AOTU 1 and L-AOTU 3 bilaterally across the two hemispheres may exist, as it has been shown that in honeybees, all compartments of the AOTUs of both hemispheres are connected via three types of TT neurons (Zeller et al., 2015). Our light microscopy data with electron microscopy reconstructions of the entire adult fruit fly brain could help to map the complete population of these visual projection neurons (Zheng et al., 2018).

| Parallel processing and bilateral integration
Parallel circuit processing is a common feature of most sensory modalities, such as the visual and olfactory sensory systems. In Drosophila, the olfactory projection neurons relaying odor information to the mushroom body and the lateral horn were proposed to mediate learned and innate olfactory behaviors, respectively (Amin & Lin, 2019;Sanes & Zipursky, 2010). The M 6 ! L-AOTU ! EB is a common pathway processing polarized light for decision-making during navigation in insects (el Jundi et al., 2014;Warren et al., 2019). Here, we found another pathway M 6 ! VMP with MV neurons relaying M 6 information directly to the VMP premotor center, where dense descending neurons link the brain to the ventral nerve cord for rapid locomotion response .
In the silk moth Bombyx mori, head orientation requires a visual

| Heading representation
A population of EB output neurons (E-PG) that is responsive to changes in head direction acts as an internal compass for navigation (Seelig & Jayaraman, 2015). The EB consists of five major types of intrinsic ring neurons (R1-R4 and P, Table 7). The R1-R4 neurons are downstream from TB neurons that connect at specific BU microglomeruli to the EB, while the P neurons connect the inferior dorsofrontal protocerebrum (IDFP) to the EB (Hanesch et al., 1989;Renn et al., 1999;Daniels et al., 2008;Lin et al., 2013;Zhang et al., 2013;Martín-Peña et al., 2014  information from the R7 neurons in both eyes (Fischbach & Dittrich, 1989). From here, only dorsal glomeruli of the four L-AOTU columns receive visual input from the posterior M 6 in both brain hemispheres, indicating that some, but not all, BU microglomeruli receive visual inputs. As the visual cue moves to the right, both the posterior and the anterior M 6 in the right brain hemisphere receive visual inputs.
From here, all glomeruli in the four L-AOTU columns in the right brain hemisphere receive visual inputs, indicating that all the microglomeruli of the right BU receive visual inputs from both the posterior and the anterior M 6 in the right eye. The difference between the two M 6 visual fields is further represented to the bilateral BUs by commissural TT neurons and ascending VBT neurons, suggesting a circuit mechanism for the heading representation during navigation.
Altogether, the morphological diversity of the visual projection neurons of the M 6 downstream circuitry suggested multi-dimensional processing of UV information via hierarchical, multiple-layer, parallel, and bilateral circuits in the Drosophila brain. Notably, the downstream circuitry, which comprises the specialized polarization-sensitive R7 neurons in the dorsal rim area (DRA) of the Drosophila's eye terminate in the medulla dorsal rim area (MEDRA), may differ from the one described herein. Furthermore, the UV-sensitive R7 and the blue-green-sensitive R8 neurons mutually inhibit each other (Schnaitmann et al., 2018). The M 6 downstream circuitry may also be involved in color vision (with an antagonistic action) for green light, as has been shown for TT neurons in locusts (Kinoshita et al., 2007;. The UV-sensitive R7 neurons also relay visual signals to the lobula via R7-Dm8-Tm5c for UV preference behavior (Gao et al., 2008;Karuppudurai et al., 2014), suggesting that the differentially processed downstream circuits of R7 neurons may convey diverse information to the central brain for color comparison in Drosophila. However, the mechanism underlying the circuit interactions between these parallel pathways (R7-Dm8-Tm5c, R7-MT-TB, and R7-MV-VBT) remains to be elucidated. Altogether, this comprehensive visual circuit map provides the basis for studying how the brain uses UV information to guide navigation behavior in Drosophila.

ACKNOWLEDGMENTS
We thank Drs. Tzu-Yang Lin and Chi-Hon Lee for their critical comments to improve the manuscript. We thank the Vienna Drosophila Resource Center and Bloomington Stock Center for the fly stocks. We are grateful to the FlyCircuit database from the NCHC (National Center for High-Performance Computing) and NTHU (National Tsing Hua University) for providing original single-neuron images.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/cne.25068.

CONFLICT OF INTEREST
The authors declare no competing interests.

AUTHOR CONTRIBUTIONS
Chu-Yi Tai, An-Lun Chin, and Ann-Shyn Chiang planned the study and analyzed the data. Chu-Yi Tai performed imaging experiments, FlyCircuit data analysis. Chu-Yi Tai and Ann-Shyn Chiang wrote the paper. Ann-Shyn Chiang supervised the project. All authors had access to all the data in this study and take responsibility for the integrity of the data and the accuracy of the data analysis.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon request.