Taking advantage of genetic methods, the present study described the morphology of candidate octopaminergic neurons in the Drosophila brain. Application of the Flp-out technique (Wong et al.,2002) enabled comprehensive visualization of the morphology of individual octopaminergic neurons. This first systematic identification revealed a common cellular architecture of octopaminergic neurons: most of them receive main inputs in one specific brain area, i.e., the posterior slope, and each neuron typically targets several defined neuropil regions in the brain. The cells differ from each other mainly in the combination of their target regions. Since OA regulates a large repertoire of behaviors, this anatomical map may help identify the underlying neural circuit for each behavioral modulation mediated by OA.
Large-scale identification of individual octopaminergic neurons
The random mosaic labeling in NP7088 and tdc2-GAL4 allowed us to visualize most of the individual GAL4-expressing neurons. Based on the double labeling of the GAL4 lines with the anti-OA antibody, we assume that 27 GAL4-expressing neurons distributed in clusters VM, AL2, VL, and ASM are octopaminergic. Among the 16 different cell types identified in the VM cluster, 11 types are OA-VUM neurons and the other five types are OA-VPM neurons. Since OA-VPM neurons are paired and some of the OA-VUM neurons seem to be duplicated (data not shown), this study identified most, if not all, of the 27 neurons in the VM cluster.
The projection patterns of the OA-immunoreactive and GAL4-expressing neurons were mostly consistent with a previous report on the cellular distribution of OA (Sinakevitch and Strausfeld,2006). Yet it remains to be confirmed whether the cells in cluster ASM and OA-AL2b2 are indeed octopaminergic (Figs. 1C, 2G, 7D). Although we reproducibly found immunolabeling in this cluster (Fig. 2G), these cells have not been reported to be OA-immunoreactive (Monastirioti et al.,1995; Sinakevitch and Strausfeld,2006). Although the crossreactivity of the anti-OA antibody to TA was controlled (Suppl. Figs. 2, 3), these neurons might be tyraminergic without producing OA, as proposed in the Drosophila larva (Nagaya et al.,2002). Despite the technical difficulty of the transmitter staining, a comparison of the expression patterns of Tdc2 and Tβh, an enzyme specifically required for OA, but not TA, synthesis, would reveal the existence of such tyraminergic neurons.
It is important to mention that both GAL4 drivers we employed in this study do not cover all the octopaminergic neurons labeled with the anti-OA antibody (Fig. 2). Since those octopaminergic cells not described in this study may play important roles in OA-mediated behavioral modulation (e.g., the neurons projecting to the ellipsoid body), completing the map by identifying the morphology of those neurons is of particular importance.
Anatomical hallmarks of octopaminergic neurons
Judged by the distribution of presynaptic and postsynaptic marker proteins, the majority of octopaminergic neurons appear to be highly polarized (Figs. 4, 5), i.e., main input and output regions are clearly separated. Varicose terminals, which are colocalized with the presynaptic marker, are distributed throughout the brain. In contrast, spiny arbors of neurons of clusters VM and AL2 are localized in the posterior slope, where the postsynaptic marker Rdl-HA is highly concentrated (Fig. 5). Consistently, OA immunolabeling was less intense in this region compared to other presynaptic regions (data not shown; Monastirioti et al.,1995; Sinakevitch and Strausfeld,2006). This polarized architecture suggests that these neurons are designed to collect synaptic inputs in the posterior slope and to modulate an array of targets by giving outputs in distinct neuropils in the brain. Afferent neurons that are presynaptic to the octopaminergic neurons remain to be identified. The characteristic polarity may open a possibility to identify the entire neural circuits underlying the neuromodulation by OA.
The single cell morphology of these neurons revealed that each cell type stereotypically innervates a distinct combination of brain regions (Figs. 14, 15). For instance, the mushroom bodies and the optic lobes are innervated by different groups of neurons. The cells of each group target distinct subregions within those neuropils (Fig. 15). For example, OA-VPM3 terminates in the fan-shaped body and the nodulli of the central complex, whereas another neuron, OA-AL2i1, projects only to the protocerebral bridge. Similarly, the six types of the descending neurons exhibit distinct projection patterns (Fig. 15). These strongly suggest that octopaminergic neurons are organized in a combinatorial manner: Each individual neuron seems to be a component of specific neural circuits. Thus, each type of octopaminergic neurons could serve as a “module” that could selectively modulate the function of a respective area of the brain, (Fig. 15). The stereotypy of these neurons might be confirmed by aligning the same type of octopaminergic neurons in different brains using computer algorithms (Jenett et al.,2006; Jefferis et al.,2007; Datta et al.,2008). However, such programs available at the current time produced large errors in aligning single cells in the entire brains of different samples (data not shown). Thus, we here present our microscopic data without such standardization (Fig. 14).
Figure 14. Two examples of stereotypy. Innervation patterns of OA-AL2i1 (A–L) and OA-VPM1 (M–U) in three different samples. The left-most column depict projections of confocal stacks illustrating the overall projection patterns of different samples. Right columns depict ramifications of each sample monitored by mCD8::GFP (white) in the neuropil staining (orange). A–L: Each sample of OA-AL2i1 exhibits a similar innervation pattern in the inferior and superior posterior slope (ipsl and spsl in B,F,J), the protocerebral bridge (pb in C,G,K), and the optic lobe (D,H,L; I me: inner medulla, lo: lobula, lop: lobula plate). The black arrow in I indicates a simultaneously labeled different cell. M–U: Each sample of OA-VPM1 exhibits a similar innervation pattern in the ventromedial protocerebrum (vmpr) and the SOG (N,Q,T). In each sample a characteristic secondary neurite was identified, which branches off ventromedial to the esophagus (oes) into te ipsilateral brain hemisphere (arrows, O,R,U). Scale bars = 25 μm.
Download figure to PowerPoint
Figure 15. Schematic diagram of the innervation areas of all individual cell types analyzed in this study. Dots represent neuropils containing varicose nerve terminals of each individual cell type. Regions with spiny ramifications are labeled with X. The target regions of individual cell types are connected by a line. Each neuropil is innervated by a distinct set of neurons, suggesting a combinatorial organization of octopaminergic neurons. VNC, ventral nerve cord; ipsl, inferior posterior slope; spsl, superior posterior slope; SOG, subesophageal ganglion; an, antennal nerve; AMMC, antennal mechanosensory and motor center; AL, antennal lobe; vmpr, ventromedial protocerebrum; vlpr, ventrolateral protocerebrum; ipr, inferior protocerebrum; spr, superior protocerebrum; LH, lateral horn; fb, fan-shaped body; no, noduli; pb, protocerebral bridge; lobe, mushroom body lobes; ca, calyx; lo, lobula; lop, lobula plate; i me, inner medulla; o me, outer medulla; la, lamina.
Download figure to PowerPoint
Comparative and functional perspectives
The overall organization and function of the brain are conserved between different insect species. Interestingly, some anatomical traits of octopaminergic neurons in Drosophila are extremely similar to those of other species, suggesting their conserved functions.
One of the morphological hallmarks of octopaminergic neurons in Drosophila is the highly enriched dendritic region in the posterior slope (Figs. 5, 15). The studies on single cell morphology of octopaminergic neurons in the honeybee and locust showed that several DUM or VUM neurons in the SOG of the locust or honeybee terminate in the corresponding region (Bräunig,1991; Bräunig and Burrows,2004; Schröter et al.,2007). However, partially due to the lack of a transgenic marker, such centralized postsynapses have been less obvious in these insects. This interspecies difference might be due to the fact that each neuromere is clearly segregated in these insects, whereas the neuromeres in the Drosophila SPG and SOG are fused and therefore condensed. Further investigation of the spatial relationship between the dendrites of different octopaminergic neurons might illuminate an internal structure within the enriched dendritic regions.
Antibody labeling of OA has visualized paired cells lateral to the midline neurons in various insect species (Stevenson and Spörhase-Eichmann,1995; Sinakevitch et al.,2005). Compared to the other insects such as honeybees, crickets, locusts, and cockroaches, Drosophila seems to have more paired octopaminergic neurons in the SOG (Stevenson and Spörhase-Eichmann,1995; Sinakevitch et al.,2005). The morphology of these paired median neurons has been unknown. The single-cell analysis in this study, for the first time, revealed the projection pattern of each subesophageal VPM neuron (Fig. 8). As in OA-VUM neurons, OA-VPM neurons, yet asymmetrically, exhibit extensive projections to distinct brain regions with similar neuronal polarity.
Importantly, our analysis identified a pair of OA-VPM neurons that might be a counterpart of an unpaired median neuron in another insect. We found that OA-VPM3 innervates the fan-shaped body of the central complex, the calyx of the mushroom body, and parts of the superior protocerebrum (Fig. 8I–L). These target brain regions match those innervated by a single subesophageal DUM neuron in the locust (SA1) (Bräunig,1991). SA1 has mirror-symmetric processes and terminate in parts of the central complex, various protocerebral regions, and the calyces. Based on the anatomical homology of the octopaminergic neurons between these two species we speculate that both neurons could fulfill similar functions. The replacement of unpaired with paired neurons might be the reason for a larger number of OA-VPM neurons in Drosophila.
In other insect species, subesophageal VUM/DUM neurons potentially modulate various sensory afferents and behaviors by innervating sensory nerves (Long and Murdock,1983; Bräunig and Pflüger,2001; Scheiner et al.,2002). Consistently, efferent octopaminergic neurons in these insects are found to project through the sensory nerves, such as the antennal nerves (Bräunig,1991; Bräunig and Pflüger,2001; Schröter et al.,2007). The neurons in the VM and AL2 clusters in Drosophila extensively innervate the terminal areas of sensory neurons including the optic lobes, the antennal mechanosensory and motor centers, the antennal lobes, and the SOG. Therefore, primary sensory information might be modulated directly by OA. Although we identified no efferent octopaminergic neuron, we assume that this might be due to a technical reason. Because our whole-mount preparation usually cuts the sensory nerves at the level close to the brain, single efferent fibers would not be detected.
The neurons of cluster AL2 innervate the medulla, lobula, and lobula plate, but not lamina (Figs. 6, 7). This cluster consists of ipsilaterally and bilaterally projecting neurons (OA-AL2i and OA-AL2b, respectively). Importantly, they share striking structural similarity to the neurons PM4 and PM5 in the locust (Stern et al.,1995; Stern,1999). They are shown to be involved in dishabituating the response of the descending contralateral movement detector. Given their homologous cellular architecture, OA-AL2i neurons in Drosophila might modulate visual information in a similar manner. It should be important to clarify whether these neurons directly synapse onto sensory neurons. Because of the lack of AL2 cluster neurons and OA-immunoreactivity in the lamina (Sinakevitch and Strausfeld,2006) (data not shown), only the inner photoreceptors (i.e., R7 and R8 cells of the ommatidia) are eligible for the direct sensory targets in the optic lobes.
Our large-scale analysis identified six different descending octopaminergic neurons: three types of VUM (OA-VUMd) neurons, one type of OA-VPM neurons, and two types of OA-VL neurons. Except for OA-VUMd3, they all have ascending processes in the brain as well as descending secondary neurites into the cervical connectives (Figs. 8, 12, 13, 15). Interestingly, all three types of OA-VUMd neurons we found belong to the posteriormost subcluster, i.e., VMlb (Fig. 10). This morphological hallmark is consistent with the organization in the locust. The majority of the descending subesophageal DUM neurons in the locust belong to the posterior subcluster (Bräunig and Burrows,2004). These descending octopaminergic neurons might regulate motor control in the thoracic ganglia, since OA is reported to stimulate certain motor behaviors, such as locomotion and grooming (Yellman et al.,1997).
In appetitive olfactory learning in the honeybee, the stimulation of the identified octopaminergic neuron VUMmx1 was shown to replace sugar reward (Hammer,1993). VUMmx1 bilaterally innervates the antennal lobes, the calyces of the mushroom bodies, and the lateral horns in the bee brain. We also found a VUM neuron in Drosophila (OA-VUMa2) that project to the same brain structures. Since OA has a similar function in olfactory learning of Drosophila (Schwaerzel et al.,2003; Schroll et al.,2006), the role of OA-VUMa2 in reward processing can now be studied.
In summary, our comprehensive anatomical analysis of Drosophila octopaminergic neurons revealed a strong overall similarity to the organization in other insect species. The conserved architecture implies importance of the development of octopaminergic neurons for their functions. In addition, Drosophila melanogaster will serve as a genetically tractable model system to study functions, development, and evolution of octopaminergic neurons.