Neuroarchitecture of the Drosophila central complex: A catalog of nodulus and asymmetrical body neurons and a revision of the protocerebral bridge catalog

Abstract The central complex, a set of neuropils in the center of the insect brain, plays a crucial role in spatial aspects of sensory integration and motor control. Stereotyped neurons interconnect these neuropils with one another and with accessory structures. We screened over 5,000 Drosophila melanogaster GAL4 lines for expression in two neuropils, the noduli (NO) of the central complex and the asymmetrical body (AB), and used multicolor stochastic labeling to analyze the morphology, polarity, and organization of individual cells in a subset of the GAL4 lines that showed expression in these neuropils. We identified nine NO and three AB cell types and describe them here. The morphology of the NO neurons suggests that they receive input primarily in the lateral accessory lobe and send output to each of the six paired noduli. We demonstrate that the AB is a bilateral structure which exhibits asymmetry in size between the left and right bodies. We show that the AB neurons directly connect the AB to the central complex and accessory neuropils, that they target both the left and right ABs, and that one cell type preferentially innervates the right AB. We propose that the AB be considered a central complex neuropil in Drosophila. Finally, we present highly restricted GAL4 lines for most identified protocerebral bridge, NO, and AB cell types. These lines, generated using the split‐GAL4 method, will facilitate anatomical studies, behavioral assays, and physiological experiments.

Understanding the core principles of operation of the central complex has been greatly enabled by the dissection of behavior at a single neuron level and the neuron-by-neuron assembly of circuits. A comprehensive anatomical atlas and genetic lines that enable manipulation of individual cell types are invaluable tools for this strategy. In this study,  Table 1 for abbreviations). The only documented roles for the NO in Drosophila are in the time course of walking activity (Strauss & Heisenberg, 1993) and in influencing handedness during locomotion (Buchanan et al., 2015). The locust neurons that connect the PB, EB, and NO and the PB, FB, and NO are sensitive to polarized light (Heinze & Homberg, 2009).
Most recently, recordings from optic-flow-sensitive neurons that connect the LAL to the NO and other neurons that link the NO to the FB in the bee have demonstrated the NO are involved in path integration (Stone et al., 2017). Finally, the fact that this structure appears to be present only in the subclass of winged insects has led to the speculation that the noduli may regulate flight (Pfeiffer & Homberg, 2014).
Structural conservation of the central complex across insect species is strong, but not absolute. The discussion that follows focuses on the anatomy of the Drosophila neuropils (Figure 1), unless otherwise noted. The PB, FB, EB, and NO are midline structures and exhibit a stratified organization (Heinze & Homberg, 2008;Homberg, 2008;Strausfeld, 1999;Williams, 1975). The PB is a handlebar-shaped structure in the posterior dorsal brain (Figure 1a,b). The EB is shaped like a torus and is tilted on its dorsoventral axis such that its ventral border defines the anterior margin of the central complex (Figure 1a,b). The FB lies between the PB and EB and represents the largest of the four central complex structures (Figure 1a-c). The bilateral noduli, historically called ventral tubercles (Hertweck, 1931) Frontal view as seen from the anterior. Neuropil masks were created in FluoRender (https://bmcbioinformatics.biomedcentral.com/articles/10.1186/ s12859-017-1694-9 or (Wan, Otsuna, Chien, & Hansen, 2009 and were aligned in the JRC 2013 brain. The color coding is as follows. PB: dark blue; FB = green; EB = orange; NO = teal; AB = rose (dorsal to NO); LAL = lilac; GA = burgundy ("shoulders" of LAL); crepine (CRE) = light blue; round body (ROB) = yellow; brain = gray. Scale bar = 20 μ. (b) Frontal (left) and sagittal (right) views of central complex neuropils. Anterior is to the right in the sagittal view. PB = purple; FB = green, EB = orange; NO = cyan; AB = rose. Scale bar = 20 μ.
(c) The same images as shown in (b), with the PB and EB removed to better visualize the location of the AB (rose). Scale bar = 20 μ.
(d) Sagittal view of one set of noduli. The dorsal nodulus, NO 1 (1), is not horizontally segmented. The medial nodulus, NO 2 , has two subcompartments, dorsal (D) and ventral (V). The ventral nodulus, NO 3 , has three subcompartments, anterior (A), medial (M), and posterior (P) complex includes the FB, EB, and "ventral tubercles" (Power, 1943), or NO. By the mid-1970s, the modern view of the central complex had emerged: Williams included the PB within the locust central complex, alongside the FB, EB and NO (Williams, 1975).
The asymmetrical body (AB) is a relatively inconspicuous structure located at the midline, adjacent to the ventral FB. It was first described in the fly as a round, almost exclusively right-hemisphere structure (Pascual, Huang, Neveu, & Preat, 2004). The AB was observed in both hemispheres in just 7.6% of 2,250 brains immunolabeled with an antibody against the Fasciclin II (Fas II) protein, which is expressed in this structure. Flies with bilateral ABs were reported to have disrupted long-term memory, leading to the suggestion that asymmetry of this structure is important for long-term, but not shortterm, memory (Pascual et al., 2004). Although thousands of GAL4 lines that drive expression in small subsets of neurons in the larval and adult fly brains have been examined (Jenett et al., 2012), the AB is the only reported instance of an asymmetric structure in the adult fly brain.
Elements with likely homology to the Drosophila AB have also been described in the grey flesh fly Neobellieria bullata and the blowfly Calliphora erythrocephala (Phillips-Portillo & Strausfeld, 2012). In both species, these bodies occur bilaterally, and one of the two is consistently smaller and less densely innervated than the other. In addition, the smaller of the two appears fragmented. Jenett et al. (2012) identified five GAL4 lines that show asymmetric innervation of the AB. Their analysis revealed lines ranging from a strong right hemisphere bias in innervation to those with asymmetric but bilateral expression, with more conspicuous expression in the right AB. This study builds on Jenett et al. (2012) by providing a systematic characterization of the neurons that target the AB, and leads us to propose that the AB be added as the fifth neuropil of the Drosophila central complex.
In this work, we present a characterization of cell types of the NO and AB, including morphology, presumed polarity, and population size.
We also generated and characterized a set of split-GAL4 lines for NO and AB cell types, reagents that will greatly facilitate functional studies. These lines are presented here in Table 2.
In addition, since publication of our description of the neurons that arborize in the PB (Wolff et al., 2015), we have gained several new insights into the PB neurons. These include (a) one new PB neuron family has been identified; (b) a neuron identified by Lin et al. (2013) has since been found in the GAL4 collection and is characterized here using the multicolor flip-out technique (MCFO; Nern, Pfeiffer, & Rubin, 2015); and (c) we generated and characterized a set of split-GAL4 lines for PB cell types; these are presented in the Results section in Table 2.
2 | MATERIALS AND METHODS

| Split-GAL4 methodology and lines
In the split-GAL4 method, the DNA-binding domain (DBD) and activation domain (AD) of the GAL4 transcription factor are separately linked to complementary leucine zippers and placed under the control of different gene enhancers to create "hemidriver lines" (Luan, Peabody, Vinson, & White, 2006;Pfeiffer et al., 2010). Each enhancer drives gene expression in an empirically determined population of neurons. In cells in which both enhancers are active, both the AD and DBD are expressed and can dimerize via their leucine zippers to reconstitute transcriptionally active GAL4 protein. The reconstituted GAL4 protein binds to the upstream activating sequence (UAS) and activates expression of the UAS-transgenes. In this case, the UAStransgenes used are "reporters" (e.g., GFP) that visually identify those cells that express GAL4.
Adult brain expression patterns of approximately 8,200 GAL4 lines, 7,000 from the Janelia collection (Jenett et al., 2012;Pfeiffer et al., 2008)    , making it possible to analyze the morphology of individual cells at high resolution using confocal microscopy. The MCFO strategy was used to identify and characterize the NO and AB cell types described here.
Briefly, the fly stocks used for MCFO contain three UAS reporter constructs (UAS-FRT-STOP-FRT-epitope tag), each with a different Cell type(s) and number of cells per hemisphere are provided for each line. One primary (best) and several good lines/cell type are provided. Some 2 lines are equally as good as the 1 lines; these are indicated in bold. a PB G2-9 .s-FBℓ1.b-NO 3 P.b&NO 3 M.b&NO 3 A.b contains PB G2-9 .s-FBℓ1.b-NO 3 P.b, PB G2-9 .s-FBℓ1.b-NO 3 M.b, PB G2-9 .s-FBℓ2.b-NO 3 A.b. b Also has weak expression in PB G2-9 .s-FBℓ3.b-NO 2 V.b. c This line targets P-EN2 cells (Dan Turner-Evans, personal communication; see Green et al., 2017, for description of cell type). The identities of the P-EN cells in the first two lines have not been characterized. Also has weak expression in PB G2-9 .s-FBℓ3.b-NO 2 V.b. d This line is stochastic-expression may occur in only one hemisphere.
epitope tag (Flag, VK5, or HA). Flp recombinase-mediated excision (Struhl & Basler, 1993) stochastically removes the STOP cassette in a small number of cells, enabling expression of one to three epitopes.
The epitope tags are subsequently labeled with appropriate antibodies (  Nern et al. (2015). Briefly, crosses were set up (day 1) and maintained at 21 C. Flies eclosed on day 14-15 and were heat shocked on day 16 (1-2 days old) by placing the vials containing the flies in a 37 C water bath for 15 min. Female flies were used almost exclusively in this study (some male flies were analyzed for the AB portion of the study) and were dissected 5-9 days following heat shock. Tissue was processed as outlined below.

| Split-GAL4 expression patterns
The expression patterns of split-GAL4 lines are visualized in Figure 2 using 20XUAS-CsChrimson-mVenus in attP18 (Klapoetke et al., 2014) as the neuronal marker and anti-BRP to label the neuropil. Expression patterns of additional lines are shown at www.janelia.org/split-GAL4 and confocal stacks corresponding to all split-GAL4 lines described here can be downloaded from that site. Instructions for requesting split-GAL4 lines can also be found at this site. To supplement the anatomy-based assignments of neuronal polarity, split-GAL4 lines were used to drive the expression of two reporter constructs, one targeted to membranes (pJFRC225-5XUAS-IVS-myr::smGFP-FLAG in VK00005); Viswanathan et al., 2015) and a second that localizes to synaptic vesicles in presynapses (pJFRC51-3XUAS-IVS-Syt::smGFP-HA in su(Hw)attP1). Synaptotagmin (Syt) is a presynaptic protein (Littleton, Bellen, & Perin, 1993). Immunolabeling was performed as described in . Additional details on reporter constructs are also described in this publication. A minimum of two female brains and three CNSs (a CNS is a brain plus ventral nerve cord, or VNC) per stable split line were immunolabeled and mounted in DPX. Detailed protocols are available at: https://www. janelia.org/project-team/flylight/protocols (see "IHC-Anti-GFP," "IHC-Polarity Sequential," and "DPX mounting"). Imaging conditions for each brain and corresponding VNC are identical, so the intensity seen in the brain and VNC for each line accurately reflects the relative levels of expression in these two regions. Microscope settings were adjusted as necessary in different lines.

| Antibody characterization
Primary antibodies used in this study are presented in Table 4 The lines shown in Table 3 correspond to those shown in Table 2. Lines are arranged numerically by primary SS line number and in ascending order. They are further arranged by neuropil region, starting with the PB, followed by the NO and last by the AB. Each text box in Table 3 includes all lines for a given cell type that are presented in Table 2. SS lines indicated in bold are the same lines shown in bold in Table 2 and identify the best lines for each cell type.
PB G1/2-9 .b-SPSi.s PB G2-9 .b-IB.s-SPS.s P P P P P P PB PB PB PB PB PB P PB PB PB PB P PB PB PB P PB PB PB PB PB B PB B B P P PB P P P G9 . .P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P P.
LAL.s-CREi.s-NO 3 P/Mi.b FIGURE 2 Expression patterns of PB, NO, and AB split-GAL4 lines in brain and VNC. Expression of 20xUAS-CsChrimson-mVenus (insertion in attP18; labeled with anti-GFP antibody, green) driven by split-GAL4 lines indicated in panels. Neuropil was visualized using anti-Brp, shown in magenta. The sparsest lines for each cell type are shown. Confocal stacks for these and additional lines noted in Table 2 can be viewed and lines can be ordered at www.janelia.org/split-GAL4. This website will be updated with additional clean lines as they become available. Scale bar = 100 μ Furthermore, they show a correspondence between GFP and endogenous BRP patterns when a panneural driver is used to drive GFPtagged bruchpilot. Brp expression is lost in brp mutants and rescued by expression of BRP in these mutants (Kittel et al., 2006  St. Louis, MO). DPX was allowed to dry for 2 days before imaging.  [SS50464]

| Image acquisition and analysis
FIGURE 2 Continued described here. On the other hand, since aligning the data to a standard brain allows a comparison between brains, registered data was used for the color depth mask searches. Details for preparation of the standard brain used in this work (JFRC2013) can be found in Aso, Hattori, et al. (2014). Neuropil masks were generated using FluoRender (software that generates three-dimensional images; https:// bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-017-1694-9 or Wan et al., 2009Wan et al., , 2012, brains labeled with anti-Brp, and a standard brain (JFRC2013). For details, refer to Aso, Hattori, et al. (2014).

| Figure preparation
In cases in which the neuron intensity in MCFO images was below the threshold required for visibility in a publication, brightness, and contrast were increased in the Janelia Workstation. In several cases, anti-nc82 label and neuron fragments from neurons that were not the focus of the figure, were "erased" using Adobe Photoshop (San Jose, CA). Brightness and contrast for each pair of brains and VNCs shown in Figure 2 were individually optimized; settings are identical for brains and corresponding VNCs within each sample.

| Identification and characterization of split-GAL4 lines
Most GAL4 lines drive expression in multiple cell types, sometimes making it difficult to isolate individual cells for anatomical analysis using the MCFO technique. Their activity in multiple cell types also makes these lines problematic for behavioral assays. To circumvent these limitations, the split-GAL4 intersectional method  in combination with optimized vectors (Pfeiffer et al., 2010) was used to reduce the number of cell types targeted in a given line.
A detailed description of this method and of the generation of the hemidriver collection of lines is provided in Materials and Methods.
Transgenic hemidriver lines (Dionne et al., 2018;Tirian & Dickson, 2017) carrying unique enhancer fragments that drive expression in cell types of interest were selected from two GAL4 collections (Jenett et al., 2012;Pfeiffer et al., 2008;Tirian & Dickson, 2017). Parental AD and DBD hemidriver lines judged to have overlapping expression patterns were crossed. A subset of these crosses is successful in producing lines in which the UAS reporter is expressed in a small number of cell types. The number of cell types that is observed in each intersection is cross-specific and highly reproducible among the progeny of a given cross. The most useful combinations were made into stable split-GAL4 lines that carry both the AD and DBD hemidriver constructs (see Methods). The "cleanest" split-GAL4 lines for each cell type (i.e., those that show expression in the least number of cell types in the brain, optic lobes, and VNC) are presented in Table 2 and  Table 2 can be accessed at www.janelia.org/split-GAL4.
The split-GAL4 lines provided in Table 2 all drive expression in fewer cell types than the original GAL4 lines, but not all are specific for a single cell type. In some cases, this is due to the limited number of AD and DBD hemi-drivers (the parent lines) in the available collections. In the case of the NO 3 M and NO 3 P cells, we were able to design many split-GAL4 lines that are expressed in both of these cell types, but we could not identify enhancers that are uniquely expressed in each of these cell types; either they do not exist or the lines in the GAL4 collection do not target them.
Three analyses were used to characterize the morphology, polarity, and population size of the cells targeted by these lines. The first, the MCFO method, is a selective approach that stochastically labels small populations of cells, providing insight into cellular morphology and neuropils targeted by a given cell type; it does not, however, reveal the population-wide projection pattern for that cell type.
Polarity data, generated in the second type of analysis in which cell type-specific split-GAL4 lines are used to drive expression of reporter constructs that target membranes and presynapses, supplements the MCFO data in that it provides molecular confirmation of the morphology-derived polarity of neurons obtained using the MCFO methodology (not shown). These polarity data rely on the correct targeting of an exogenous fusion protein and can overpredict presynaptic sites in cases where expression levels are too high (Aso, Hattori, et al., 2014). Although the expression levels of this protein are titrated  in this system, caution should still be exercised in interpreting these patterns. Similarly, while the MCFO anatomical data also has its limitations, it can reliably identify the predominant arbor type. Since neither of these methods provides the accuracy that can be achieved using electron microscopy (EM), and since arbors often have both inputs and outputs, EM-level resolution will be required to establish the true percentage of spines and boutons for each terminal.
In the third method of analysis, brains were prepared using the Chrimson-Venus marker to reveal the expression patterns ( Figure 2 and www.janeliaorg/split-GAL4) of the GAL4 lines listed in Table 2.
These samples reveal both the approximate population size of a given neuron type as well as each cell type's projection pattern. We observed minimal brain-to-brain variability in the number of cells in some lines. The variability is due to loss of cell bodies during dissection, stochasticity in gene expression, difficulty in resolving individual soma when they were tightly clustered, and perhaps some variability in neural composition between individuals. The cell counts provided are therefore not precise population sizes. Taken together, these three labeling techniques provide a comprehensive description of each cell type analyzed.

| Nomenclature
The nomenclature system used here includes three features that unambiguously identify a neuron so that it can be recognized by its name: the structures targeted by the neuron; the primary neurite's path relative to the midline; and the polarity, deduced from the anatomy of the arbor. A detailed description of this nomenclature system can be found in Wolff et al. (2015); several points are reiterated here to assist in interpreting the neuron names presented in the results. The abbreviations "i" for ipsilateral, "c" for contralateral, and "ic" for instances when a projection arborizes in both the ipsi-and contralateral counterparts of a structure, are appended to all but the first neuropil in the neuron's name to indicate their location relative to the first neuropil. For example, in the hypothetical neuron "A-Bi-Ci-Dc," the arbors in neuropils B and C are on the same side of the midline as A's arbor, the primary neurite crosses the midline between neuropils C and D, and the arbor in structure D is in the contralateral hemisphere.
It is generally possible to distinguish between the two basic arbor morphologies of a neuron at confocal microscope-level resolution, although EM studies will be required to quantitatively describe all arbors. These two morphologically distinct arbors, which resemble thin fibers and blebs, provide insight into the neuron's polarity and direction of information flow: the fine, thin arbors are typically associated with functional input whereas the bleb-like boutons are typically recognized as the functional output of a neuron. These anatomical features are included in the nomenclature system used here.
The abbreviations "s," for terminals that appear to be predominantly thin spines, "b" for terminals that appear to consist predominantly of boutons, and "s.b" for obviously mixed terminals, are appended to each structure in the neuron's name. In the hypothetical neuron A.s-  may not necessarily coincide with functional differences. The arbors of some neurons extend beyond these nebulous boundaries. For example, some neurons discussed below have extensive arbors in the LAL and a few fine terminals from these arbors extend minimally into the adjacent crepine (CRE). In an effort to both avoid confusion and simplify the names of neurons with such minimal excursions into adjacent neuropils, the name of the neuropil with minor representation is not included in the neuron's name. A comprehensive anatomical description that includes these details is, however, included in the text.

| Recommended abbreviations for neuron names
The nomenclature used in this publication and in Wolff et al. (2015) is comprehensive but too cumbersome for reiterative use in papers and talks, so a set of standardized abbreviations is presented in Table 5. In an effort to keep the abbreviations as brief as possible, only critical information is retained. For example, FB layers and NO designations (1, 2, and 3) are excluded from the PB-FB-NO contractions since compartment information, such as anterior, dorsal, etc., uniquely identifies these cell types and is included. This abridged nomenclature is consistent with, and builds upon, a loose convention established by several labs working on central complex neurons, and it evolved from discussions among several individuals working on these neurons (see Acknowledgments). The following rules were used to derive the abbreviations in Table 5. 2. The remaining neuropils follow the three-letter designations presented in Ito et al. (2014). To avoid confusion with the single letter notations of sequential neuropils in the abbreviation (see rule 5), only the first letter of the three-letter code is capitalized (e.g., Slp, which, if SLP, could be confused for an unknown neuropil "S," followed by the LAL, followed by the PB).
3. The field has established P-EG and E-PG as the abbreviations for the neurons PB G1-9 .s-EBt.b-D/V GA.b and PB G1-8 .b-EBw.s-D/V GA.b, respectively. These abbreviations will remain as such.

Neuropil subdomain information is included only when necessary
to distinguish between otherwise identical abbreviations. In these cases, the first letter of the subdomain's name is appended as a subscripted capital letter. For example, the canal P-EG, PB G1-9 .s- (Note the original nomenclature system did not list input neuropils first, but the gall tip cell is, in fact, an E-PG cell.) In these two cases, the subscripted "C" and "T" disambiguate these two cell types from the P-EG and E-PG noted in #3, above.
5. Apparent input (spiny morphology) neuropils are listed first followed by neuropils with mixed input and output followed last by apparent output neuropils (boutons).
6. Input, mixed, and output neuropils are separated with hyphens.
7. Variants that are not morphologically distinct are assigned arbitrary, nonsubscripted numerals that follow the designated abbreviation (e.g., P-EN1 and P-EN2). overlaid on a box plot. While there is a great deal of variability in AB volume both between brains and between hemispheres, the left AB is consistently on average 25% of the volume of the right AB, but ranged from 10 to 53% 8. Cells with formal names that convert to pronounceable acronyms will adopt those acronyms (e.g., Delta7).

| NO neurons
The classification of a neuron as a PB, FB, EB, or NO cell is somewhat arbitrary since central complex neurons arborize in one to three structures of the central complex and could therefore be classified by any one of the three targeted neuropils. We began our studies of the central complex with the PB (Wolff et al., 2015), so all neurons that were found to arborize in the PB were assigned to the PB class, even though many of them project to additional structures within and outside of the central complex. Neurons presented here as "NO neurons" exclude both those described previously as PB neurons that also arborize in the NO, for example, the PB-FB-NO and PB-EB-NO neurons (Lin et al., 2013;Wolff et al., 2015), as well as tangential FB neurons that arborize extensively throughout one layer of the FB and also send more restricted projections to the NO (for example, see Hanesch, Fischbach, & Heisenberg, 1989). In the discussions that follow, "NO neuron" refers exclusively to those neurons described in the results.
This work identified as many as nine distinct "nodulus neurons"

| NO 2 -targeted neurons
The NO 2 -targeted neurons identified to date project to both the dorsal and ventral subcompartments of the medial nodulus, NO 2 . This distribution is implied in the names of each of these neurons by the absence of the "D" and "V" (for dorsal and ventral) designations. Five NO 2 cell types are described, although four are rare. Although the sample size for these four neurons is small, their morphologies differ significantly from the most common cell type, LAL.s-NO 2 i.b (below), and they display features that distinguish them from one another.
These morphological criteria suggest that these four neurons are strong candidates to be bona fide distinct neuronal types; they are therefore described here.
LAL.s-NO 2 i.b The ipsilateral LAL.s-NO 2 i.b neuron has predominantly spiny branches in the LAL whereas the NO 2 arbor is a densely packed bundle of boutons (n > 52; Figures 3b and 4b). The LAL arbor is more proliferative than the NO 1 neuron's LAL arbor. Its dorsal boundary often creeps slightly into the adjacent CRE (Figure 4b). The arbor's intrusion into the CRE is minimal and therefore not included in the neuron's name.

Four putative cell types
Four additional "putative" NO 2 cell types are described below. Each of these possible cell types has been seen just one or two times unless otherwise noted, all in the same split-GAL4 line, SS04682, so we cannot be certain that they represent bona fide cell types rather than cells with aberrant projections. If this is the case, it suggests the commonly seen LAL.s-NO 2 i.b cell is highly prone to errors in line SS04682 and these putative cells may represent developmental mistakes. However, this possibility seems less likely since the morphological features of these cells are robust and elaborate-not typical of the projections we have occasionally seen in other lines that resemble developmental miswiring. Alternatively, the scarce number of sightings may simply reflect that these neuron types are rare or infrequently targeted with

| The asymmetrical body is a distinct, bilateral neuropil that is asymmetric in size and bilaterally innervated
The analysis presented here indicates that the AB, while asymmetric in morphology, is present in both hemispheres in all brains (n > 167;  (Figure 5d). Given the asymmetry in size, the term "asymmetrical body" still appropriately describes this neuropil.

| AB neurons and neuron families
Three AB neurons or neuron families have been identified and are Population-wide view of tile and canal cells. (a1) SS02191-driven expression of a membrane-targeted epitope (blue) and a presynaptically-targeted epitope (red). Line SS02191 shows expression in only the tile cell: PB G1-9 .S-EBt.B-D/V GA.B. Asterisks identify cell bodies; red asterisk identifies primary neurite that is missing its cell body, which are sometimes lost in dissection (red syt signal seen in the cell bodies is likely due to protein trapped in the golgi, as is frequently seen with exogenously expressed proteins). Scale bar = 20 μ. (a2) Cells in PB are labeled with a membrane marker (blue) and reference channel is identified with α-nc82 label (gray). All 18 glomeruli of the PB are targeted by this cell type. Scale bar = 20 μ; also applies to a3 and a4. (a3) The radius (left) and depth (right; sagittal view of EB) of the EB arbor are shown; the reference channel is included to identify the EB boundary. The arbor does not extend to the canal and populates only the posterior third of the EB (anterior is to the right in sagittal view). (a4) Both compartments of the gall are densely packed with boutons. White line delineates the boundary between the dorsal and ventral gall compartments. α-nc82 label to the right of the gall is not part of the gall. (b1) SS02195-driven expression of the canal cell: PB G1-9 .S-EBc.B-D/V GA.B. Markers are the same as described in Figure 12a1 legend. Cell bodies were identified in the confocal stack and their locations are indicated by asterisks in this maximum intensity projection, where they are difficult to resolve. Faintly stained ramifications belong to an unrelated cell type (arrow). Scale bar = 20 μ. (b2) The canal cell also populates all 18 glomeruli of the PB. Cells are labeled with a membrane marker, PB labeled with α-nc82. Scale bar = 20 μ; also applies to (b3) and (b4). (b3) The canal cell EB arbor is offset toward the canal (left) and extends the depth of the EB (right, sagittal view, anterior is to the right). (b4) Also in contrast to the tile cell, the canal cell gall arbor does not fill the dorsal and ventral domains. The ventral gall arbor clearly wraps around the ventral gall. Refer to MCFO images in Figure 13 to better visualize the distribution of the dorsal gall arbor FIGURE 13 The canal cell: PB G1-9 .s-EBc.b-D/V GA.b. (a1) Line SS02198 shows expression in both the canal and tile cells and enables a direct comparison between the two cell types. Shoots from the EB arbors of the two green canal cells enwrap the EB canal (yellow asterisk; see also a2) whereas such filamentules are notably absent from this red tile cell (and most tile cells). Note the distinct difference in density of the gall arbors: the red tile cell arbor has more boutons than the sparse canal arbors (green asterisks). G3, G4, and G9 identify the glomeruli of the PB that are targeted by these cells. Scale bar = 20 μ [from line SS02198]. (a2) Footprints of the tile (red) and canal (green) cells differ in radius (top, frontal view) and depth (bottom, sagittal view). Both cell types occupy the posterior shell of the EB; the canal cell projects even deeper, into the medial and sometimes even the anterior shells, although only a few boutons reach the anterior shell. Anterior is to the right in bottom panel. (a3) Canal cell gall arbors (green) are sparser than tile cell gall arbors (red) and generally even less dense than in this sample, for example, as seen in Figure 13b,d. (b) A canal cell that targets G4 in the PB. As noted for the tile cell, PB arbors from canal cells can also spill into neighboring glomeruli, shown in the right inset, in which G9's arbor extends fine, sparse shoots into G8 (asterisk). The canal cell was named for the filamentules that project from the EB arbor toward the central canal (top right). Sagittal view of the EB (lower right) illustrates that the bulk of the canal arbor is concentrated at the posterior of the EB, with a few branches reaching more anteriorly (anterior is to the right). Note that this cell, which targets an even-numbered PB glomerulus, projects to just the tip of the ventral gall and sends a small projection more ventrally (arrows). The canal cell gall arbor is not as robust as the tile cell's gall arbor in that it does not fill the gall. The dorsal gall is in a different plane of view than is shown in the inset and does not receive input from this cell [from line SS02195]. (c) The distribution of the canal cell's EB arbor is variable; in some cells, boutons almost completely enwrap the canal (G5 and G6, in this figure), in others, a uniform swath extends the radius of the EB (G7), and in still others, the greatest density of boutons is offset toward the canal (G1). Several additional examples and the depth to which they penetrate the EB are shown in the lower panel; only the orange cell (G6) reaches the anterior shell, as seen in the side view (anterior is to the right). The projections of these cells to the dorsal/ventral gall are shown in (d) [SS02195, SS02198, SS04776]. (d) The profile of the canal cell's gall arbor is distinct from that of the tile cell but like the tile cell, it follows the odd/even glomerulus to D/V gall projection rule. D = dorsal gall. V = ventral gall. (d1) The mint green G6 cell from (c) projects to the ventral gall (V) in panel d1 but rather than filling this neuropil, as does the tile cell (e.g., red cell in Figure 13a3), it appears to track along this neuropil's surface [SS02198]. (d2) The red G1 cell shown in (c) also exhibits this tracking behavior in the gall. This three-panel series from posterior (top panel) to anterior illustrates that this red arbor snakes between the dorsal "family." All members of the family share several features in common: they all arborize in the AB and FBℓ8, the FB arbors are columnar, and the AB terminals are mixed, consisting of both boutons and spines.
The presence of spines in the AB of this family is unique among the three AB cell types characterized to date and identifies them as the only output pathway (based on morphology) so far identified from this neuropil. Family members differ in the number of arbors in FBℓ8 and in whether or not they arborize in FBℓ1. The suggestion that this cell type's morphology is not highly stereotyped, but instead exhibits a conspicuous degree of variability (in other words, that these are all variations of the same cell type), is supported by the observation that the various forms occur in each of the GAL4 and split-GAL4 lines examined, as well as in both split half parents of a given split-Gal4 line. For example, a range of morphologies has been seen in VT006486 and two split-GAL4 lines for which VT006486 is a parent line, SS04457 and SS04423. It is possible that results from behavior and physiology assays, as well as type-specific split-GAL4 lines, will indicate these forms are distinct cell types.
Arbors of these family members are predominantly spiny in the SLP and the remaining arbors are composed primarily of boutons and are therefore presumed to be output. The SLP.s-ABic.b-FBℓ8.b neuron, the "core" member of the family, is shown in Figures 7c and   9a. The remaining projections in cells in this family seem to exhibit some level of stochasticity, as there is a great deal of variability in the layers of the FB that are targeted. Some cells in this family project neurites that terminate in boutons in various combinations of layers 1, 2, 4, and 5 of the FB, and one bouton was seen in the EB. Typically, just one to several layers receive a bouton or two. The neuron shown in Figure 9b strongly resembles that shown in Figure 9a, but in addition to the relatively robust arbors in the SLP, FBℓ8, and AB, it also extends a few boutons into FB layers 1, 4, and 5 (e.g., red asterisk).
The arbor in FBℓ8 in this family always spans the midline but varies in the degree to which it permeates this layer; as few as five medially located boutons have been seen. The cells shown in Figure 9c

| PB neurons
Since publication of Wolff et al. (2015), several advances have been made in our understanding of neurons that target the PB. First, we have identified several new PB neurons that have features in common with a published cell type, the octopaminergic OA-AL2i1 cell (Busch, Selcho, Ito, & Tanimoto, 2009). In addition, new insight has been gained into a previously described neuron, PB G1-8 .s-EBt.b-D/V GA.b (Wolff et al., 2015). Finally, a neuron discovered by Lin et al. (PB 1 glomerulus ->EB C -IDFB DSB (Lin et al., 2013) has now been identified in GAL4 lines, enabling additional characterization of this cell type, which we call PB G1-9 .s-EBc.b-D/V GA.b to be consistent with our previously described nomenclature system. The latter two cells, PB G1-8 .s-EBt.b-D/V GA.b and PB G1-9 .s-EBc.b-D/V GA.b, share many features in common, as described below. In addition, split-GAL4 driver lines for the PB neurons have been generated and are presented in Table 2; these lines are generally highly specific for a particular cell type.

| A family of PB neurons with morphologically similar PB arbors
We have identified a number of PB neurons that share features in common with one another and with a published PB octopaminergic neuron, the OA-AL2i1 cell (Busch et al., 2009); see Figure 10  , the green cell (originates from G2) targets the ventral gall (V). Lower panel, the orange cell (originates from G6) targets the ventral gall (V). The GA arbors appear to mostly "surround" the gall compartments more than fill them. This is particularly evident in the right panels of the lower series, in which the ventral gallseeking cell does not fill the VGA (gray neuropil) but instead tracks along its anterior surface. The white lines delineate boundary between dorsal and ventral gall compartments. Scale bar = 20 μ [SS04776] The combination of the intricate morphology of the cells and crowded GAL4 lines has made it difficult to label individual cells in unique colors, so the descriptions and images that follow are best approximations of the neuronal morphology.
Although all the projections could not be unambiguously traced, there is clearly a wide scope of neuropils targeted by this collection of neurons. Their projections proliferate throughout the brain and some appear to send projections to the optic lobes (Figure 10a-c) and the ventral nerve cord (VNC; not shown). The common feature among these neurons is a sinuous arbor of variable density that courses through the PB (Figure 10d1-d4), suggesting that some PB functions may be modulated by octopamine. In some neurons, the PB arbor traverses just half of the PB (Figure 10b, inset). In other instances, it appears to run the full extent of the PB (Figure 10a, inset; 10c, inset; 10d1-d4), but since there may be two labeled neurons in these specimens the limits of single PB arbors cannot be definitively determined.
Since the anatomy of these cells is not well-defined in our images, we have not named them.
3.6.2 | A revised account of the "tile cell," PB G1-9 .s-EBt.b-D/V GA.b The functional and anatomical unit of the PB is the glomerulus (G), which derives its name from the Latin "glomus," meaning "ball of yarn." In keeping with the standard established in Hanesch et al. (1989) for Drosophila, glomerulus is used here rather than the generic term "slice," which refers to a transverse division of a structure (Ito et al., 2014). The glomeruli are numbered G1-G9, from medial (G1) to lateral (G9). The specific glomeruli targeted by a given cell type are indicated following the neuropil designation PB. The EB is segmented along both its transverse and longitudinal axes. Two distinct transverse volumes are revealed by the arbors that fill them and were termed tiles (abbreviated "EB.t") and wedges, based on their geometry (Wolff et al., 2015). A third volume, the "canal," is introduced here.
We identified a cell that bears a strong morphological resemblance to the tile cell, a PB-EB-GA cell (Wolff et al., 2015), in seven split-GAL4 and several generation 1 GAL4 lines. Two significant differences suggested it was a new cell type, but the analysis outlined below indicates they represent the same cell type.
We originally reported that the tile cell targets only eight of the nine glomeruli, skipping the most lateral glomerulus, G9, and that its EB arbor, characterized in GAL4 line R33A12, is confined to the posterior ring of the EB. In contrast, the apparently new cell targets all nine glomeruli (Figures 11a1,a2,b1,b2 and 12a1,a2). Furthermore, a small subset of the new cell's EB arbors project tendrils that reach toward the central canal of the EB (Figure 11b1, green asterisk; compare to neighboring red arbor that lacks these tendrils, red asterisk).
Each of these filamentules terminates in a single bouton (Figure 11b1, inset, arrow). The number of bouton-capped tendrils varies from cell to cell, from zero, as was seen with the original tile cell in line R33A12, to several (Figure 11b1 inset).
In other respects, these two cells are indistinguishable. First, their PB arbors are both spiny and fill their resident glomeruli, often spilling into a neighboring glomerulus (Figure 11b2, red arrow). Second, both EB arbors are confined to the posterior shell (Figures 11a3,c2 and 12a3 right). And third, the GA arbors consist primarily of boutons, which fill either the dorsal or ventral subcompartment of the GA, depending on the glomerulus of origin (Figures 11c3,c4 and 12a4; cells that arborize in the odd glomeruli navigate to the dorsal gall and those that target even glomeruli track to the ventral gall, as described in Wolff et al. (Wolff et al., 2015). (not shown), confirming the two variants are not only one and the same type, but also that this class of neurons arborizes in all nine glomeruli. Furthermore, the three SS lines that include VT040589 as a split half parent (SS02191, SS27853, and SS16814), all drive expression in the EB tile cell in all nine glomeruli (e.g., SS27853, Figure 11). This cell's name is therefore being changed from PB G1-8 .s-EBt.b-D/V GA.b to PB G1-9 .s-EBt.b-D/V GA.b to reflect its presence in G9 (note the change in the name is from G1-G8 to G1-G9).
3.6.3 | The canal cell, PB G1-9 .s-EBc.b-D/V GA.b, strongly resembles the tile cell Since publication of Wolff et al. (2015), MCFO performed on additional GAL4 lines revealed a PB-EB-GA cell not reported in that study.
Also in Wolff et al. (2015), an effort was made to correlate the PB cell types described in that report with PB cell types described in Hanesch et al. (1989) and Lin et al. (2013); see Table 3A We now believe the "canal cell," not the tile cell, is equivalent to the PB 1 glomerulus -> EB C -IDFB DSB cell. (Currently, there is therefore no Lin et al., 2013 correlate for the tile cell.) Since MCFO enables a detailed analysis of circuitry and morphology, here we build upon the description provided in Lin et al. (2013) and compare the canal cell to the morphologically similar tile cell.
The canal cell's name is derived from a prominent anatomical feature of this neuron: its EB arbor partially enwraps the canal, or central hole, of the ellipse (Figure 13). Following the nomenclature system adopted in Wolff et al. (2015), this cell's formal name is the PB G1-9 .s-EBc.b-D/V GA.b cell, where the "c" that follows "EB" stands for canal.
The canal and tile cells share many features in common, including overall morphology, putative input and output regions, and circuitry.
First, light-level analysis reveals that the canal cell has primarily boutons in the EB and gall (".b" for boutons; Figure 13a-d) and predominantly spiny arbors in the PB (".s" designates a spiny morphology; Figure 13b, right inset), as described for the tile cell. Furthermore, the PB arbors for both cells often occupy more than one glomerulus, spilling over into the neighboring glomerulus ( Figure 13b, right inset, asterisk). Occasionally, the arbor completely fills two neighboring glomeruli and a small portion of a third (not shown). Second, the canal cell also targets all nine glomeruli of the PB (Figure 12b1,b2). Third, canal cells also respect the odd/even glomerulus:D/V GA wiring pattern ( Figure 13c,d for several examples). Fourth, the circuitry between individual glomeruli and the corresponding regions in the EB is identical between the two cell types [not shown; see Figure 14b from (Wolff et al., 2015) for visual depiction of this circuitry].
The key differences between the tile and canal cells are the distribution and morphology of the EB and GA arbors. The canal cell EB arbor projects shoots of variable length around a portion of the circumference of the canal, partially enwrapping the canal (Figure 13a1, a2,b,c). Relative to the tile cell, the bulk of the canal cell EB arbor is shifted from the periphery to the center. The density of boutons in the canal EB arbor and the area over which the arbor is dispersed from periphery to center varies significantly from cell to cell (Figure 13a1,a2,green arbors;13b,top right: 13c). Consequently, in some instances it is difficult to distinguish the tile and canal cells from one another. The depth of the EB arbor also differs between the two cell types.
Whereas the tile cell EB arbor is confined to the posterior ring ( Figure 12a3, right), the canal cell EB arbor occupies not only the posterior ring, but also invades the medial and sometimes even the anterior ring of the EB (sagittal views in Figures 12b3 and 13a2,b,c).
The gall arbors in the tile and canal cells are usually distinct, although sometimes they, too, are hard to differentiate. The tile cell's gall arbor is robust and fills the entire volume of the dorsal or ventral gall (Figures 11c3,c4 and 12a4). In contrast, the canal cell's gall arbor is much sparser (Figures 12b4 and 13a1, green asterisks; 13a3, green arbor; 13b, left inset and arrows; 13d) and appears to either creep along the surface of the gall or snake between the dorsal and ventral compartments of the gall (Figure 13d1,d2). Sometimes it seems to hover at the dorsal or ventral tip (Figure 13d3, pink cell).
While the variability in morphology can make it difficult to distinguish between the tile and canal cells at the single cell level, distinctions between the two cell types are unmistakable when the entire population of each cell type is labeled in split-GAL4 lines specific for each cell type (compare tile cell in Figure 12a to canal cell in Figure 12b). These images confirm the MCFO-based observations noted above, as follows. First, all nine glomeruli are targeted by both cells (Figure 12a1,a2,b1,b2). Second, the tile cell EB arbor does not extend to the canal (Figure 12a3, left), it fills only the posterior shell of the EB (Figure 12a3 right; anterior to the right), and the arbors in the dorsal and ventral gall are dense and uniform (Figure 12a4). In contrast, the canal cell's EB arbor exhibits a significant concentration around the canal (Figure 12b1,b3 left) and is dispersed throughout the posterior, medial, and anterior rings of the EB (Figure 12b3, right), and the gall boutons are not densely concentrated throughout this neuropil, instead appearing to be concentrated dorsal to the dorsal gall and around the periphery of the ventral gall (Figure 12b4).

| DISCUSSION
There is increasing evidence that many complex insect behaviors, such as navigation and sleep, rely crucially on the central complex, yet our understanding of this key brain region is in its infancy-even the fundamental building blocks of its neuropils are, for the most part, uncharacterized. With the current tools available in Drosophila, comprehensively cataloging these neurons and generating the tools to manipulate them is a logical and feasible first step, the results of which will facilitate a more sophisticated understanding of the circuit computations and the behaviors controlled by neurons of the central complex. With a catalog of neurons, a set of genetic driver lines that target each of these cell types and a battery of behavioral assays through which to run these targeted fly lines, the tools exist to quantify the behavioral effects of manipulating the activity of defined, small subsets of neurons. Such experiments provide a way to infer the involvement of different neuron types and neuronal networks in specific behaviors. The work described here provides such a descriptive catalog of neurons and genetically targeted GAL4 driver lines for two structures, the noduli and the asymmetrical body, as well as GAL4 driver lines for PB neurons. The availability of these tools will facilitate studies of the roles of the NO, AB, and PB and the neuronal cell types that populate these brain structures.
4.1 | The asymmetrical body: The fifth central complex structure The neuropils considered to be constituent components of the central complex have changed over the decades. Power (1943) defined the neuropils of the central complex to include the "central body, the ellipsoid body and the pair of ventral tubercles." The central body (Flögel, 1878), included what became known as the ellipsoid body (Lowne, 1892) and the fan-shaped body (Dietl, 1876); reviewed in (Strausfeld & Seyfarth, 2008). The noduli were known as the ventral tubercles. More recently, the central complex has been defined as "a group of modular neuropils across the midline of the insect brain" (Pfeiffer & Homberg, 2014), "…interconnected neuropils and nuclei that populate the midline of the forebrain-midbrain boundary region" (Strausfeld & Hirth, 2013), and "a system of interconnected neuropils lying at, or about, the midline of the protocerebrum" (Ito et al., 2014).
Although the modular architecture of the central complex structures is conspicuous (e.g., the glomeruli of the PB and the trajectory patterns of neurons that project to, from, and within the central complex structures), it is the assigned boundaries that encompass the central complex that seem to be the feature that defines these structures as members of the central complex.
Here, we illustrate that the Drosophila AB, which appears to be a structure that is distinct from the FB, meets the criteria outlined above for central complex neuropils: It is a midline neuropil; it falls within the boundaries defined by Power (1943) and Strausfeld and Hirth (2013); and it is interconnected (to the FB and SLP) by a network of previously undocumented (with one exception) neurons.
Since the AB meets all the criteria previously used to define neuropils as components of the central complex, we propose that the AB be added as a fifth neuropil of the central complex of Drosophila.
The AB is not unique to Drosophila. Phillips-Portillo and Strausfeld (2012) describe the presence of likely homologous bilateral, asymmetrically sized ABs in N. bullata and C. erythrocephala. Their work also identifies a tangential FB neuron that bears a resemblance to the SLP-AB-FBℓ8 neuron described here. It remains to be determined if the AB is more widely represented in other insect orders.

| A mostly asymmetric neuron
The right AB is significantly larger than the left. At a minimum, this difference is likely due to a combination of smaller arbors in the left AB and the lower frequency with which the left AB is targeted: only the ipsilateral-contralateral-projecting form of the SLP-AB neuron, which arborizes in both the left and right ABs, targets the left AB, whereas the ipsilateral and contralateral-projecting forms of the SLP-AB neuron target exclusively the right AB. Thus, the right AB appears to receive a disproportionately larger share of information from the SLP, although the right and left hemispheres appear to be equally represented as sources of input. Notably, this left-right bias is restricted to the AB, as a parallel preference is not shown for the SLP. The availability of genetic lines that target AB-specific cell types will enable experiments aimed at revealing the relevance of this left-right bias.

| Unusual properties of the NO neurons
Three unusual features distinguish the five most commonly seen NO neurons described here from other central complex neurons. First, in contrast to the majority of PB neurons described to date, the projections of four of these five NO neurons are ipsilateral. Second, while anatomical features identify distinct input and output neuronal populations in other central complex neuropils, the noduli appear to be sites for receiving primarily input from other neuropils (boutons appear to be the predominant anatomical feature in the noduli in confocal micrographs of NO, PB-FB-NO, and PB-EB-NO neurons). Golgi preparations and data from the likely locust equivalent of the PB-FB-NO neuron (the CPU4 neuron), however, indicate these NO arbors are mixed (Hanesch et al., 1989;Heinze & Homberg, 2008;Stone et al., 2017); perhaps the intensity of the dense populations of boutons masks the presence of spines in confocal preparations. Third, although the noduli do receive input from central complex neuropils (e.g., via the PB-FB-NO and PB-EB-NO neurons, from FB tangential neurons, etc.), the majority of direct input for this new set of neurons is provided by just one neuropil, the LAL. Such a restricted thoroughfare of communication is in stark contrast to the PB neurons, for example, which have a much broader and more diverse network of direct communication.
The LAL.s-CREc.s-NO 3 Pc.b cell type is distinct from the other four common NO neurons in that it delivers contralateral, rather than ipsilateral, input from the LAL and CRE to NO 3 P. The posterior compartment of NO 3 is therefore unique in that it is the only nodulus subcompartment to communicate directly with the contralateral hemisphere. Given that NO 3 P (and NO 3 M) also receives ipsilateral terminals from the LAL via LAL.s-CREi.s-NO 3 P/Mi.b, this subcompartment may act as a limited integration center between the fly's left and right sensory fields.

| Potential roles of the NO neurons
Physiological data from two neurons in the sweat bee offer insight into a likely role for the NO neurons described here. The TN1 and TN2 neurons ("noduli tangential neurons") share a high degree of anatomical homology with the LAL-NO neurons: TN1 and TN2 are ipsilateral neurons with input branches in the lateral central brain and blebbed branches in the noduli (Stone et al., 2017). Recordings from these two cells reveal they fire in response to simulated backward and forward flight, respectively, and that the rate of firing is dependent on the stimulus velocity, suggesting these neurons encode speed using optic-flow and can thereby track the distance traveled by the bee.
Similar physiological features and path integration functions would not be unexpected for the apparent homologous Drosophila neurons.
The LAL is the primary source of input for the NO neurons described here and its activity may provide additional insight into the roles of the LAL-NO neurons. It is a large, bilateral neuropil that is highly interconnected with neuropils of the central complex. Functionally, the LAL is considered to be a sensorimotor integration center, based on several lines of evidence in various insect species. For example, in crickets and moths, activity in LAL neurons is associated with walking (Iwano et al., 2010;Mishima, 1999;Zorovic & Hedwig, 2013).
In the locust, assorted LAL neurons exhibit changes in activity in response to various aspects of flight, implicating this brain region in flight control (Homberg, 1994). In Drosophila, LAL neurons involved in walking backwards have been documented (Bidaye, Machacek, Wu, & Dickson, 2014).
It has been suggested that the noduli are involved in walking and motor control in Drosophila (Buchanan et al., 2015;Strauss & Heisenberg, 1993). The neurons implicated in left-right turning bias in locomotion are the PB-FB-NO neurons (Buchanan et al., 2015), which have presumed input (fine terminals) in the PB, and presumed output (boutons) in the FB and NO. The authors speculate that the bias to turn in one direction or the other is influenced by an interplay between the nodulus subdomains that are targeted by the different PB-FB-NO cell types. Direct communication between the PB-FB-NO neurons and the LAL-NO neurons is not unexpected, as Stone et al. (2017) have shown synaptic contacts between the bumblebee equivalents of these two cells, the CPU4 and TN cells, respectively.
Considering the sensorimotor contribution made by the LAL in various types of movement, the LAL-NO neurons described here are strong candidates to contribute to the circuits involved in turning.

| NO neurons: An incomplete picture
The catalog of NO neurons described here is incomplete. Our analyses of other GAL4 lines have identified several large-field FB neurons that also arborize in the noduli, as well as other brain regions that we are currently characterizing; some of these neurons are illustrated in the Golgi stains of Hanesch et al. (1989). Hanesch et al. (1989) also describe other cell types with arbors in the noduli that have so far eluded our identification. Finally, it seems likely that there would be output neurons from the NO, although we have not yet identified such neurons in our studies. Electron microscopic-level analysis should provide a path to identifying these neurons.

| Defining "cell type" and neuropil boundaries
The debate continues to swirl over what constitutes a distinct cell type. Morphology and function have long been accepted as reliable criteria to distinguish cell types. While morphology is a straightforward and easy means of classifying cell types, it can be misleading in that cells that appear identical may have functional differences. For example, Green et al. (2017) describe clearly distinct physiological roles for two PB neurons that appear to have indistinguishable morphology at the light level. Morphological features evident with light microscope-level resolution will therefore likely be insufficient to distinguish all cell types, so knowledge of some combination of synaptic connectivity, functional properties and the genetic programs used to specify these attributes will be necessary to fully define cell types.
Similar limitations confound the assignment of neuropil boundaries and subcompartments. Synaptic density varies considerably across brain regions and this variation has provided landmarks used to define the neuropils of the fly brain (Ito et al., 2014). While the boundaries of some structures are unambiguous (e.g., the PB and EB), neuropil margins are not universally so clear-cut, with many neuropils appearing to meld seamlessly with adjacent neuropils. The opportunity to map the domains of arbors within neuropils identifies distinct regions that are not revealed by differences in synaptic density (e.g., wedge and tile domains in the EB). For example, the mushroom body lobes can be divided into a series of nonoverlapping compartments with distinct functions by the extent of the arbors of dopaminergic input neurons and mushroom body output neurons (Aso, Hattori, et al., 2014;Aso, Sitaraman, et al., 2014). The LAL provides an example of one neuropil that may have functionally distinct subregions. It is a large neuropil with no obvious boundaries revealed by anti-Brp staining, yet the arbors of many neurons that target this neuropil exhibit strong regional preferences. Mapping the domains of these arbors may identify regions that are functionally distinct. 4.7 | Obtaining a comprehensive description of cell types, circuits, and the tools to manipulate them Three major efforts aimed at cataloging all the neurons in the Drosophila brain are in progress. One, typified by this and others' work (Aso, Sitaraman, et al., 2014;Tuthill, Nern, Holtz, Rubin, & Reiser, 2013;Wu et al., 2016), characterizes one structure at a time using light microscopy in combination with the generation and analysis of highly specific GAL4 driver lines. The second method is a modern implementation of the Golgi approach of randomly labeling small numbers of neurons in order to describe their morphology (Chiang et al., 2011). And the third, which is now becoming practical at the required scale, involves reconstruction of neuronal morphology and circuits through analysis of image volumes collected using electron microscopy (see, for example, Takemura et al., 2017, Takemura et al., 2013.
We believe that such light and electron microscopic-level analyses will be highly synergistic. Light microscopy, with genetically marked cells, provides the ability to observe the morphology of hundreds of individual cells of the same cell type in many different individuals, providing insights on stereotypy. However, its dependence on GAL4 drivers means that completeness of coverage cannot be assured. Conversely, electron microscopic analysis, while usually limited to a single sample, not only ensures completeness but also enables visualization and quantification of synaptic connectivity. Moreover, since EM samples do not carry transgenes expressing ectopic membrane proteins that can interfere with development, wiring errors may be less likely. While only electron microscopy is likely to provide the complete wiring diagram of a circuit, light level analysis of genetic driver lines will be needed to provide the critical bridge between circuit maps and the tools required to precisely manipulate the activity of their individual components.

| ACKNOWLEDGMENTS
The authors are grateful to Laszlo Tirian and Barry Dickson for providing Drosophila lines (the "Vienna collection") prior to publication and to Jens Goldammer, Masayoshi Ito, Ryo Minegishi, and Aljoscha Nern for generously sharing MCFO data. We thank A. Nern for data that was used to generate panels in Figure 10. The first iteration of the rules for neuron abbreviations described here came from a brain- Core, who sets up crosses, members of Janelia Fly Light, who do the dissections and immunohistochemistry, and members of the imaging team, who image the fly brains. Scientific Computing supports the Janelia Workstation, which is used to view the confocal stacks. We thank V. Jayaraman for insightful discussions and Kazunori Shinomiya for his input on the ROB abbreviation. We are grateful to Chris Doe, Ulrike Heberlein, V. Jayaraman, Nick Strausfeld, and two anonymous reviewers for critical and valuable comments on the manuscript.