The term `connectome’ refers to the mapping of all neural connections within an organism’s nervous system or a confined part of it. These `wiring diagrams’ can be defined at different levels of scale, corresponding to levels of interest or the spatial resolution of imaging, for example, the microscale, mesoscale and macroscale [53]. A connectome at the macroscale (light microscope level) attempts to resolve different brain regions or neuropils and the pathways in between; these brain maps were established over the last hundred years for various species. These days with the help of various new techniques and increased computing power, the meso- and microscale (electron microscope) levels come into focus. At the mesoscale level, the morphology of distinct populations of neurons within a processing unit (for example, a column or a neuropil) is mapped. This level of analysis can be complemented by the microscale level, which involves mapping single neurons and their connectivity patterns (synapses), which according to Sporns et al. [53] will remain infeasible for an entire brain, at least for the near future. Recently, two ambitious scientific research projects, the Human Brain Project (by the European Union) [54],[55] and the BRAIN Initiative (by the United States) [56],[57], were launched to map these connection patterns in the human brain.
At the meso- and microscale levels, the basic architecture of sensory neuropils in both vertebrates (for example, the visual cortex in the human brain [58]) and invertebrates (for example, the optic lobes of the compound eyes in insects and crustaceans [43],[59]) is characterized by columns and layers. The vertical columns, for example, in the insect lamina and medulla [11],[12] are composed of repetitive subsets of afferent fibers (for example, those of the retinula cells) and characteristic postsynaptic neurons (for example, monopolar cells) that form the basic functional unit of a system (for example, visual system). Often, these columns are horizontally layered (for example, strata M1 to M6 in the medulla).
In the present study, we analyzed the pycnogonid visual neuropil at macro-, meso- and microscale levels to examine the principles that underlie this (simple) visual system and whether they compare to more complicated ones.
In the low-resolution stack, the macroscale observations of Lehmann et al. [49] can be confirmed. After entering the brain, the fiber bundle with the R-cell axons is split; one part of the axons ends in the first visual neuropil, and the other part passes the first visual neuropil and terminates in the second.
At the mesoscale level, aside from the R-cells, six different cell types can be distinguished in the first visual neuropil: five descending and one ascending cell type. The neuron gestalten are identified with two different approaches, providing support that both our three-dimensional-reconstruction and the Golgi-profiles give correct pictures of the neurons.
Three types of descending cells (D1, D2 and D3) are responsible for the subdivision of the first visual neuropil into two hemineuropils; these cells do not cross the border in between. In contrast, D4 neurons occur in both hemineuropils at once and provide lateral interactions between the two hemineuropils. The interpretation of the D5 cells is difficult. Here, these cells are found only in the right hemineuropil, which is most likely a sampling artifact, and the D5-cell bodies of the left hemineuropil are beyond the examined volume and, hence, are not reconstructed. At the microscale level, the D-cells are frequently postsynaptic to the R-cell axons and hence are second-order neurons. One individual R-cell is presynaptic to several D-cells and one individual D-cell is postsynaptic to several R-cells, indicating divergence and convergence. Concerning the synaptic pattern, no reliable separation between the five different D-cells could be made in the high-resolution stack. However, the reconstructed cells vary in the tangential size of the field they cover in a way that is analogous to their appearance in the medium-resolution stack, indicating that the synaptic pattern is similar in all descending cells.
The ascending neurons are higher-order neurons of a wider field throughout both hemineuropils. These cells are commonly presynaptic and sometimes postsynaptic to R-cells and, hence, play a feedback role in the system.
Furthermore, at the mesoscale level, it is observed that the first visual neuropil is split into two hemineuropils or columns. This is visible in both the SEM images and the three-dimensional reconstructions. The most plausible explanation of this subdivision is that one hemineuropil is linked to the anterior and the other to the posterior eye of the ocular tubercle. Additionally, in the two hemineuropils, at least three different layers of similar thicknesses are observable. In the upper third of the neuropil, the neurites of the unipolar cells enter the neuropil. Here, just a few collaterals were found. In the medium range of the neuropil, a number of things happen: most of the collaterals of the unipolar cells are found here, the branching and bifurcation of the D2 and D3 neurons occurs in this region, and finally the D4 neurons build here their tangential branches that reach into the two hemineuropils. Furthermore, in the medium range of the neuropil, which is analyzed at the microscale level in the high-resolution stack, additionally various synapses occur (whether and where synapses occur in the upper and lower ranges of the neuropil remains unclear at present because these regions were not studied at higher resolution). In the lower third of the neuropil, no more branching or bifurcation occurs, but numerous collaterals are found.
This analysis reveals that the R-cells provide the input into the system, primarily on the D-cells. Because the D-cells rarely appear to be presynaptic in the first visual neuropil, these cells most likely synapse and, hence, integrate information to higher visual centers that were not identified in this study. These centers could be the second visual neuropil or the arcuate body, which in chelicerates is closely associated with the visual system [60]. The A-cells play a special role in this system, being pre- and postsynaptic to both R- and D-cells. Hence, these cells collect information from the input (R-cells) and the second-order cells (D-cells) but also circulate information back to these cells. Mechanisms such as lateral inhibition, contrast enhancement and other filter functions could be behind this feedback loop. Furthermore, principles of divergence in the R-cells and convergence in the D-cells are found. A summary of the visual pathways are given in the wiring diagrams in Figure 8.
A comparison of our findings with those in other arthropods proves to be difficult, as representatives of only a few taxa have been studied in sufficient detail to allow comparison of neuron morphology. Especially for median eye visual systems, just a few Golgi studies are available.
Hanström [61] reported for Limulus that neurites with cell bodies around the neuropil enter the median eye neuropil. Some of these neurites end in the arcuate body and some below the arcuate body. Clear statements on the morphology of these cells are lacking, but their position is the same as the descending unipolar cells found here. Strausfeld et al. [62] reported ascending broad field L-cells in the first median eye neuropil of Cupiennius salei (Araneae) that spread through a roughly circular area equivalent to several R-cells. By comparison, the ascending cells of A. langi also spread through wide reaches of both hemineuropils. The three-dimensional-EM study by Lacalli [63] of the larval nauplius eye center of the copepod Dactylopusia sp. is quite revealing. Here, the three eyecups of the nauplius eye are connected to the naupliar eye center. This neuropil is subdivided into three cartridges, each receiving R-cell axons from one of the three eyecups. Several second-order unipolar neurons (LR-cells) with cell bodies above the neuropil postsynaptic to the R-cell axons are found. Additionally, higher-order neurons (M- and E-cells) occur in the neuropil. A similar subdivision (two eyes, two hemineuropils) is found here in the first visual neuropil of A. langi. The morphology and synaptic pattern of copepod LR-cells is similar to that of the pycnogonid D-cells, but cells presynaptic to the R-cells, similar to the A-cells in pycnogonids, have not been identified.
The only arthropod visual system studied in great detail so far is that of the lateral compound eyes in some insect and crustacean species, namely three-dimensional-TEM of Drosophila[11]–[13],[64], Golgi- and Golgi-EM-studies of insects [43],[65]–[68], and Golgi- and Golgi-EM-studies of crustaceans [69]–[73]. The lamina’s (that is, first visual neuropil’s) cell types are best characterized in the fruit fly Drosophila melanogaster, but the principles are similar in other insect species. The R-cells 1 to 6 provide input from each ommatidium and synapse to the lamina cartridges, the functional units of the lamina, which are composed of approximately 13 cells: the processes of five monopolar cells (L1 to L5), one or two amacrine cells, as well as three medulla neurons (C2, C3 and T1) and three glial cells. Additionally, two types of long visual fibers from the ommatidium, R7 and R8, pass the lamina and project to the medulla (second visual neuropil) [7]. In contrast, in crustaceans, R-cells 1 to 7 end in the lamina and R8 in the medulla. Here also, monopolar cells are found with similar characteristics as in insects. However, there is some disagreement about their number and nomenclature [69],[74],[75].
The synaptic organization in the lamina of Drosophila is studied and reviewed in detail by Meinertzhagen and O’Neil [7] and by Meinertzhagen and Sorra [11]. In the lamina, the R-cells are predominantly presynaptic to the monopolar cells L1 to L3 and to the amacrine cells. The L-cells in turn have only a few presynaptic sites (to R- and other L-cells) in the lamina. The amacrine cells are frequently presynaptic to R- and L-cells and often to T-cells. Finally, of the medulla neurons, only in C-cells do a few synapses occur, being presynaptic to L-, T-, and amacrine cells; T-cells are free of synapses in the lamina. All of these synapses are often multiple-contact synapses (dyads, triads, and tetrads). In the lamina of the crayfish Pacifastacus leniusculus the R-cells are also presynaptic to the monopolar cells [71].
