Anatomy of the lobula complex in the brain of the praying mantis compared to the lobula complexes of the locust and cockroach

Abstract The praying mantis is an insect which relies on vision for capturing prey, avoiding being eaten and for spatial orientation. It is well known for its ability to use stereopsis for estimating the distance of objects. The neuronal substrate mediating visually driven behaviors, however, is not very well investigated. To provide a basis for future functional studies, we analyzed the anatomical organization of visual neuropils in the brain of the praying mantis Hierodula membranacea and provide supporting evidence from a second species, Rhombodera basalis, with particular focus on the lobula complex (LOX). Neuropils were three‐dimensionally reconstructed from synapsin‐immunostained whole mount brains. The neuropil organization and the pattern of γ‐aminobutyric acid immunostaining of the medulla and LOX were compared between the praying mantis and two related polyneopteran species, the Madeira cockroach and the desert locust. The investigated visual neuropils of the praying mantis are highly structured. Unlike in most insects the LOX of the praying mantis consists of five nested neuropils with at least one neuropil not present in the cockroach or locust. Overall, the mantis LOX is more similar to the LOX of the locust than the more closely related cockroach suggesting that the sensory ecology plays a stronger role than the phylogenetic distance of the three species in structuring this center of visual information processing.

the behavioral aspects of prey recognition and prey catching, relatively little is known about the neuronal machinery underlying stereoscopic distance estimation and prey capture.
The compound eyes are specialized for stereopsis in several ways.
The eyes are large, forward-directed and consist of about 9,000 ommatidia (Kral & Prete, 1999). As shown in Tenodera australasiae, the total field of view of both eyes covers all directions except a small spot in the neck region (Rossel, 1979). Frontally, both eyes show a considerable binocular overlap of about 708 (Rossel, 1979(Rossel, , 1986. The frontal flattening of the eyes results in a reduction of the inter-ommatidial angle in the frontal eye region to less than 18 and has thus been considered a visual fovea (Rossel, 1979).
How visual signals from the two eyes are integrated in the brain to enable stereopsis, however, is unknown. Like in other insect species, visual signals are processed in three distinct neuropils in the optic lobe: the distal lamina, the medulla, and the proximal lobula complex (LOX; Leitinger, Pabst, & Kral, 1999;Strausfeld, 2012). The lamina and the medulla are layered neuropils with retinotopic organization, similar to their arrangement in other insect species. In contrast, the mantis LOX is compartmentalized into several distinct substructures, unlike its organization in many other insect taxa. Whereas the LOX consists of a single neuropil in bees, two distinct neuropils, termed lobula and lobula plate, are present in flies, butterflies, and beetles, while multiple nested neuropils have been distinguished in locusts and grasshoppers (Ito et al., 2014). The internal organization of the LOX in a praying mantis was first described by Cloarec (1968). She distinguished four distinct subunits in Mantis religiosa. Likewise, Leitinger et al. (1999) described four subunits in Tenodera sinensis, while Strausfeld (2012) recognized three LOX subdivisions with retinotopic organization.
Work in flies showed that the lobula and lobula plate serve distinct roles. The lobula plate is involved in global motion vision with topographic organization of cardinal motion directions represented in four layers (Borst, Haag, & Reiff, 2010;Borst & Helmstaedter, 2014) whereas the lobula serves a role for small target detection, visual fixation, and figure-ground discrimination (Aptekar, Keleş, Lu, Zolotova, & Frye, 2015;Lin et al., 2016;Nordstr€ om & O'Carroll, 2006;Trischler, Boeddeker, & Egelhaaf, 2007). In contrast, only little information is available on parallel processing of visual information in the LOX subunits in locusts (e.g., regarding polarization vision, Homberg et al., 2011) and no data exist for praying mantises. Although a variety of motion-sensitive neurons were characterized in the LOX of M. religiosa (Berger, 1985) and Tenodera aridifolia (Yamawaki & Toh, 2003), the authors did not identify the specific arborization domains of these neurons in the LOX. To provide a basis for functional studies on stereoscopic vision in praying mantises, we have analyzed the anatomical organization of centers for visual processing in two mantis species, focusing on the LOX because of its enigmatic structure and key position in visual information processing. In addition, we compared the organization of LOX subunits with those of two related insect species, the Madeira cockroach and desert locust, through three-dimensional neuropil reconstructions and g-aminobutyric acid (GABA) immunostaining.

| Animals
Female adult praying mantises (Hierodula membranacea, Rhombodera basalis) were obtained from colonies of Katharina W€ ust (M&m W€ ust, M€ uhlheim Germany) and from a mantis stock at Newcastle University.
The animals were kept at temperatures between 228C and 308C.
gregaria (Homberg, Vitzthum, M€ uller, & Binkle, 1999  3.2 | Comparison of the LOX in the praying mantis, locust, and cockroach by means of synapsin staining showed sparse immunostaining without further layering. The three distalmost layers (1-3) were less distinct than in the mantis or locust (Figure 6f). In the locust and the mantis, the accessory medulla was virtually devoid of immunostaining (Figure 6a,b) whereas in the cockroach is was densely supplied by GABA-immunoreactive processes ( Figure 6c). In all three species, immunostained fibers in the second optic chiasm connected the medulla to the LOX (Figure 6g-i). In the locust, these neurites were exceedingly fine (Figure 6h).  (Figures 6g and 7d). In addition, processes from tangential neurons, connecting the LOX with the central brain, entered layer I at its (a-c) Horizontal sections through the medulla at the level of the accessory medulla (AME). In all three species immunostained somata are scattered in the anterior soma rind (aS) near the first optic chiasm (OCH1) and in the posterior soma rind (pS) adjacent to OCH2. The AME is virtually free of immunostaining in the mantis (a) and locust (b), but shows dense immunostaining in the cockroach (c). l 5 lateral; p 5 posterior. (d-f) Horizontal sections through the medulla illustrating the distribution of immunostaining across medulla layers. The ten layers (1-10 from distal to proximal) in the ME of S. gregaria conform to previous studies (e.g., Beetz et al., 2015). Layering in H. membranacea (d) and R. maderae was adjusted to the layering scheme of S. gregaria for easier comparison. px 5 proximal. immunostaining showed that the LOX in the optic lobe of the praying mantis can be partitioned into 5 distinct modules, four of which receive retinotopic input. As judged by relative position, internal organization, and connectivity to other LOX subunits or brain areas, three of these subunits, the OLO, ALO and DLO, have obvious homologs in the desert locust and Madeira cockroach. In addition, one subunit in the locust, the ILO, and one subunit in the praying mantis, the SLO, were not identified in the other species and thus seem to be taxon-specific.
In the central brain, the central complex, a brain area receiving prominent visual input for spatial orientation (Pfeiffer & Homberg, 2014) is large and highly differentiated. In contrast, the anterior optic tubercle, providing massive visual input to the central complex in bees, ants, locusts, and butterflies could not be recognized in the mantis brain.

| Medulla layers and LOX subunits in the mantis, locust, and cockroach
The layering of GABA immunostaining in the medulla showed striking similarities between the locust and mantis but appears to be much less differentiated in the cockroach (Figure 6). A particular feature in the mantis was a highly differentiated layer 4 which contrasted against a rather uniformly stained layer 4 in the cockroach and locust. Most dramatic differences were observed in immunostaining of the accessory medulla, which as demonstrated in the cockroach, houses the internal circadian clock of the insect (Stengl & Arendt, 2016). Whether these differences relate to the different activity phases of these insects (cockroach: nocturnal; locust, mantis: diurnal) will have to await further studies.
In all three insects, a distally located, retinotopically organized LOX subunit could be identified which we named outer lobe (OLO). The OLO faces the medulla and receives direct retinotopic input from the  (Figure 8b). Thus, the locust OLO consists of the same number of layers as found in OLO1 and OLO2 of the praying mantis taken together, suggesting that the locust OLO corresponds to the segregated OLO1/OLO2 aggregate in the mantis. However, some differences exist with regard to the relative intensity of GABA immunolabeling in the four OLO layers between locust and mantis. While immunostaining in the most distal layer (layer I of OLO1) is particularly intense in both species, the most proximal layer (layer II of OLO2) is invaded only sparsely by immunoreactive processes in the mantis, but considerably more densely in the locust (Figures 7b and 8b). The cockroach OLO was recognized as a single layer in synapsin staining, but 2 layers were distinguished with GABA immunostaining. As in the mantis and locust, GABA immunostaining is strongest in the most proximal layer (Figure 9d).
The DLO could be identified in all three insect species based upon its unique position and connectivity. Another common feature was its strong GABA immunostaining, originating from fibers entering the DLO at its proximal end. The DLO has at least three layers in the praying mantis, revealed by GABA immunostaining but is differently segregated into two parts in the cockroach illustrated by synapsin staining (Figure   5f). In the locust, the DLO is undivided and neither in the locust nor in the cockroach could we identify a stratification within the DLO.
The most proximal LOX neuropil, the ALO, likewise appears to be homologous in the locust, cockroach and mantis. In all three species, the ALO was partly continuous with the superior lateral protocerebrum. In the praying mantis, it is divided into a dorsal and ventral unit. Leitinger et al. (1999) outlined two anteriorly located LOX neuropils in the praying mantis T. sinensis that they referred to as Lo3 and Lo4.
They probably correspond to ALO-D/DLO and ALO-V in this study immunostaining. These layers were not discernable in GABA immunostaining, but most probably correspond to the four layers that were also detectable in synapsin immunostaining of the locust LOX.
Two LOX neuropiles are unique to the praying mantis and the locust, respectively. These are the stalk lobe (SLO) in the mantis (see below) and the ILO in the locust. The ILO flanks the ALO from posterior and is very weakly GABA-immunoreactive ( Figure 8). It exists as a clearly segregated LOX module only in the locust but not in the mantis or cockroach. Except for connections with the OLO and ALO (Elphick, Williams, & O'Shea, 1996;Homberg, 2002;Homberg et al., 2004;), its connectivities to other brain areas and its functional significance have not been resolved. The praying mantis has its own unique LOX neuropile, the proximally located, tunnel shaped SLO.
We did not find a corresponding structure in the cockroach nor

| Relating LOX modules of mantis, cockroach, and locust to those of the fly
The strongly segregated LOX structure of the praying mantis deviates from the insect optic lobe ground pattern proposed by Strausfeld, 2009Strausfeld, , 2012 (Meyer et al., 1986). In all five insect taxa, the praying mantis, the locust, the cockroach, the fly and the bee, the most distal of those layers is stained strongly and thus probably contains a high concentration of GABA. Strausfeld refers to part of the cockroach LOX, which might correspond to the DLO, as lobula plate (figure 4.1 in Strausfeld, 2012). The lobula plate in flies receives retinotopic input from the medulla via uncrossed fiber bundles. Retinotopic input from the medulla also exists for the mantis DLO (Figure 2d), but at present, there is no evidence for an internal retinotopic organization of the DLO. The fly lobula plate does not only receive parallel input via the medulla but additionally is provided with uncrossed, retinotopic input from the lobula. We found retinotopic input via uncrossed fibers from the OLO to the ALO in the praying mantis ( Figure 2c). Thus, there is the possibility that the fly lobula plate corresponds to the DLO and/or ALO in the praying mantis. However, as for the DLO we were not able to identify retinotopy within the ALO and, therefore, these conclusions remain speculative until corresponding cell types have been identified, for example, via intracellular recordings.
An alternative explanation for the existence of a high number of LOX constituents is that some of them derive from modules located within the central brain in other insect species as was suggested by Elphick et al. (1996) and Strausfeld (2012). These modules could be laterally displaced optic glomeruli. Optic glomeruli receive converging input from ensembles of retinotopic lobula output neurons and are thought to process distinct features of the visual surround (Mu, Ito, Bacon, & Strausfeld, 2012;Strausfeld, Sinakevitch, & Okamura, 2007). Aggregating certain modules involved in processing visual information of particular importance within the OL could improve information processing by shortening travel distances of electrical signals in feed forward and recurrent neuronal connections.

| Parallel visual processing in the mantis LOX
The presence, arrangement and connectivity of neuropils in the LOX of the praying mantis suggest that parallel as well as sequential processing of visual signals from the medulla takes place. Behavioral studies show that praying mantises can detect, fixate, and track visual objects by head movements keeping the objects in an acute zone of highest spatial resolution (Prete, 1999;Rossel, 1979) but also perform optomotor responses to large field visual motion (Nityananda et al., 2015). Prey is identified by a combination of visual cues including overall size, contrast to background, location in the visual field and apparent speed (Prete, 1999). Distances are estimated through motion parallax induced by side-to-side movements (peering) at larger distances (Poteser & Kral, 1995) and through binocular disparity in the near range (Nityananda et al., 2016;Rossel, 1983Rossel, , 1986).
This illustrates that object-background discrimination is an essential task of the visual system, as well as specific binocular interactions for distance perception. In contrast, color vision and polarization detection have not been demonstrated and circumstantial evidence indicates that mantises may be monochromatic (Rossel, 1979;Towner & Gaertner, 1994).
How the different LOX subunits contribute to these performances, is not known. Intracellular recordings from LOX interneurons combined with morphological identification of the recorded neuron were achieved by Berger (1985) in M. religiosa. He characterized motionsensitive neurons responding to a small moving disc, bars, and grating stimuli. Many neurons have a strong preference for small moving objects directly in front of the animal, however some LOX neurons also respond to large field motion as was also found by Yamawaki & Toh (2003). Although the innervated subunits of the LOX were not identified by Berger (1985), all neurons had tangential ramifications in distal areas that might correspond to the OLO or ALO. Many of these neurons had side branches apparently in the DLO or other unidentified proximal regions of the LOX. Axonal projections were in the ipsi-or contralateral protocerebrum as shown for the neurons in Figure 4 of this study. The TOpro1-neuron ( Figure 4a) shows high similarity to the nondirectionally motion sensitive L7 cell recorded by Berger (1985).
The remaining three cell types of this study were newly identified.
Neurons with arborizations in both OLs, as found by Berger (1985) and in this study are promising candidates for being involved in binocular vision.
The current study establishing five distinctly organized subunits in the mantis LOX paves the way for future studies unraveling the distinct functional role of specific LOX subunits in visual tasks.

ACKNOWLEDGMENTS
We thank Drs. Erich Buchner (University of W€ urzburg, Germany) and Timothy G. Kingan (University of Arizona, USA) for gifts of