Due to the difficulties in collecting Remipedia, data on their brain anatomy so far were limited to the studies by Fanenbruck et al. [15, 16]. Fortunately, we were able to obtain three specimens of Remipedia preserved for immunocytochemistry. In this study we present the first analysis of the brain anatomy of Remipedia based on immunocytochemical methods and laser-scanning microscopy. These methods provide a much clearer resolution of neuropils, tracts, and single soma clusters than previously available [15, 16], and therefore allow for a more detailed description of neuroanatomical structures. Besides new insights concerning the general brain anatomy, our comprehensive description of the 5HT-IR in the brain resolves individually identifiable cell groups in Remipedia.
The neuroanatomical differences we detected in S. tulumensis and G. frondosus might reflect a phylogenetic divergence between the representatives of two remipede families, Speleonectidae and Godzilliidae. However, it is equally conceivable that the differences are the result of the histological fixation procedure under field conditions. Furthermore, a circadian fluctuation of the serotonin level as known from other crustaceans (see below) could be responsible for the differences of 5HT-IR in both species.
Our results have led to some differences in the interpretation of the brain anatomy of Remipedia compared to Fanenbruck et al. [15, 16]. These authors described a small unpaired transverse midline neuropil in the posterior proximity to the chiasm of the OGT that they identified as the central body (Figure 1D). Fanenbruck et al. characterized a neuropilar structure that most likely corresponds to the protocerebral bridge anterior to the chiasm of the OGT. In this position we identified the 5HT-ir midline domain sd2. Due to the branching pattern of the 5HT-IR in this domain and its medial position in the protocerebrum we interpret sd2 as the central body, which is positioned anteriorly to the chiasm of the OGT (Figures 4E and 5D). Furthermore, the 5HT-ir domains 4–6 are located within the protocerebral bridge, and sd9 most likely corresponds to the lateral accessory lobe of the central complex (Figure 5E).
Fanenbruck et al. [15, 16] described a small glomerular neuropil located anterior to the ON as olfactory satellite neuropil. We detected two small domains (sd12 and sd13), both receiving input from antenna 1. These domains seem to belong to the same neuropil that is innervated by projections of two 5HT-ir neurons (sE1; Figure 6C). Because of the connection to antenna 1, we conclude that this structure resembles compartments of the LAN2, and that it is not an olfactory satellite neuropil, as previously suggested [15, 16].
The dominant neuropils in the deutocerebrum of Remipedia are the ONs. We counted about 300 spherical olfactory glomeruli for Godzilliognomus frondosus and approximately 375 olfactory glomeruli for Speleonectes tulumensis. In both species, all of these olfactory glomeruli are innervated by approximately eight 5HT-ir neurons from the deutocerebral soma cluster D (Figures 6E, E’ and 8F, F’). An innervation of the olfactory glomeruli by 5HT-ir neurons is known from a variety of Pterygota, Collembola, Decapoda (Malacostraca) and Cephalocarida; however, the number and branching patterns differ between species [33–36]. All representatives of Pterygota studied to date show one to three 5HT-ir neurons innervating the antennal lobe, with additional projections into the protocerebrum . In Collembola, one centrifugal neuron situated in the subesophageal ganglion innervates one or four olfactory glomeruli of the antennal lobe, but never all olfactory glomeruli . In the studied Crustacea that possess 5HT-ir neurons associated with the ONs, these neurons always innervate the ipsilateral ON and lack projections into the protocerebrum. The most comprehensive information is available for Decapoda (reviews [35, 37]). Two different types of 5HT-ir olfactory interneurons were classified by Johansson : the giant interneurons and the smaller globuli cells. Apart from the identifiable dorsal giant interneuron, several other large 5HT-ir neurons were observed with species-specific variation in number, e.g., two to four pairs in spiny lobsters and crayfish per brain hemisphere (e.g., [39–41]). Furthermore, high numbers of 5HT-ir globuli-type cells, which are much smaller than the giant interneurons, were documented for a variety of decapods, reaching numbers up to several hundred cells . Several authors showed that the immunoreactivity of some 5HT-ir neurons is restricted to certain subregions within each olfactory glomerulus, e.g., the cap and base in the American lobster  or the subcap and base in the spiny lobster . Additionally, an interconnection by 5HT-ir neurites between neighboring olfactory glomeruli has not been observed . Data for non-decapod Malacostraca concerning 5HT-IR are sparse. Moreau et al.  detected diffuse staining in the olfactory glomeruli in Mysidacea, with the possible origin in two cell bodies adjacent to the ON. In Amphipoda, a giant tritocerebral 5HT-ir neuron innervates most parts of the brain, including the ONs .
Stegner and Richter have recently investigated the brain anatomy of a representative of Cephalocarida using immunocytochemistry . This group takes on an important role in the discussion of the phylogenetic position of Remipedia, because a close relationship to Remipedia has been proposed by several molecular studies [18–20]; [45–47]. In Hutchinsoniella macracantha, the only species of Cephalocarida for which data on brain anatomy exist to date, the olfactory glomeruli are arranged in seven vertical columns . Two 5HT-ir neurons innervate the ONs. However, one of the seven columns is not innervated by 5HT-ir neurites. In the other six, the innervation is concentrated on the inner margin of each olfactory glomerulus. Adjacent olfactory glomeruli, the shape of which is very different to that of other Mandibulata as will be discussed below, are interconnected by 5HT-ir neurites .
Besides Cephalocarida, only a few non-decapod Crustacea have been investigated for 5HT-IR in the deutocerebrum. These include Branchiopoda, Copepoda and Mystacocarida, all of which lack 5HT-ir interneurons [28, 30, 31, 48].
In contrast to Decapoda or Cephalocarida, each olfactory glomerulus in Remipedia seems to be innervated relatively homogenously by 5HT-ir neurites that are not restricted to certain subregions within the olfactory glomeruli (Figure 6F). Serotonin immunoreactive neurites, directly linking adjacent olfactory glomeruli as in Cephalocarida, are absent. Remipedia do not possess any projections from the 5HT-ir cells into the protocerebrum as do the corresponding neurons in many insect taxa . In summary, Malacostraca, Cephalocarida, Hexapoda and Remipedia show 5HT-ir innervation of the glomeruli in the ON, but the number, arrangement and innervation pattern of the 5HT-ir neurons differ among these taxa. At this point, we cannot decide whether the 5HT-ir innervation of the ONs is part of the ground pattern in Tetraconata or a functional convergence, an issue that has recently also been discussed with regard to 5HT-IR of chelicerate chemosensory pathways . A basic serotonergic innervation may be a feature that characterizes all neuromeres in the central nervous system of Euarthropoda, so that the deutocerebral serotonergic system may just be a modification of such a basic supply .
Organization of the olfactory neuropils
Within the glomerular organisation of the ONs known from numerous arthropod taxa, differences in number, shape and arrangement are evident. As in Remipedia, spherical olfactory glomeruli occur in hexapods like Diplura, Zygentoma , Collembola  and most Pterygota , but also in malacostracans, such as Stomatopoda , Leptostraca (Kenning M, Müller CHG, Wirkner CS, Harzsch S, unpublished results; ) and marine Isopoda . Many decapod crustaceans possess wedge-shaped or cone-shaped olfactory glomeruli, while Archaeognatha seem to be the only hexapod taxon exhibiting elongated olfactory glomeruli (review  and references therein; ). Even though the detailed shape of cephalocaridan olfactory glomeruli was difficult to resolve by immunocytochemical methods, it is certainly not spherical . Studies on Branchiopoda, Copepoda and Mystacocarida revealed that these taxa do not possess ONs and consequently do not have any glomerular structures in the deutocerebrum [30, 31, 48].
Olfactory glomeruli of spherical shape were proposed to be part of the tetraconate ground pattern . Recently, spherical olfactory glomeruli were described in some chilopods , which led to the suggestion that spherical olfactory glomeruli might be part of the mandibulate ground pattern. In this view, elongated, wedge- or cone-shaped olfactory glomeruli represent a derived attribute. However, because elongated or drop-shaped olfactory glomeruli also occur in certain representatives of Chilopoda , it is as yet unclear which shape represents the plesiomorphic character state as part of the mandibulate ground pattern.
The glomeruli in the ONs of Remipedia are arranged in at least four sublobes that are clearly separated from each other (Figure 6F). From the center of the ONs, neurite tracts extend into these different sublobes. Within the sublobes, the olfactory glomeruli are arranged like clusters of grapes. Subdivision of the ONs into several lobes is also known from a certain subgroup of Decapoda, the Coenobitidae (terrestrial hermit crabs). However, in contrast to Remipedia, their wedge-shaped olfactory glomeruli are arranged radially around the outer margin of three sublobes and the sublobes are not as clearly separated as those in the Remipedia (e.g. [29, 32]). In conclusion, the anatomy of the ONs in Remipedia represents a special and unique way of arranging olfactory glomeruli without any close resemblance to the topology of other Crustacea or Hexapoda.
Connections between Deuto- and Protocerebrum
Distinct fiber tracts, the olfactory globular tracts (OGTs), that connect the deutocerebral ONs to protocerebral structures were reported for Cephalocarida [33, 55], decapod Malacostraca [25, 56], Copepoda  and Remipedia ([15, 16], this study). In Hexapoda, one or more tracts with equivalent topology, here called antennocerebral tracts, project from the antennal lobe to specific protocerebral neuropils (reviews [35, 58]). Protocerebral targets of the antennocerebral tracts in the Hexapoda are the mushroom bodies. The OGTs target the multilobed complex in the Cephalocarida, the hemiellipsoid bodies (HEs) and the medulla terminalis (termed lateral protocerebrum; [25, 59]) in Malacostraca, and in Remipedia the HEs. In the majority of studied Malacostraca and in Remipedia, the paired branches of the OGT form a characteristic chiasm in the center of the protocerebrum [15, 16, 25]. Recently, a chiasm of the OGT similar to that in Remipedia and Malacostraca was reported for the copepod Tigriopus californicus.
In Remipedia, we detected a small neuropil (Figure 4D) that lies near the HE and is connected to it by 5HT-ir neurites. From a 5HT-ir domain sd11 in this neuropil, some 5HT-ir axons project into the protocereral bridge (Figures 5E and 7). This resembles the situation in the decapod Cherax destructor. Based on the connectivity to the HE and the pattern of 5HT-IR, we interpret this neuropil as a medulla terminalis, which together with the HE forms the lateral protocerebrum. In Mandibulata, the lateral protocerebrum is associated with the optic neuropils. The four retinotopic optic neuropils, lamina, medulla, lobula plate and lobula , which characterize the central visual pathway associated with the compound eyes of malacostracan Crustacea and Pterygota, are absent in Remipedia . Therefore, the medulla terminalis receives no input from the optic neuropils, and this might explain the reduction of the medulla terminalis in Remipedia.
Serotonin immunoreactivity in the neuropil of the HE has been detected in the decapod Procambarus clarkii and Orconectes rusticus, but not in Cherax destructor and Pacifasticus leniusculus. In Remipedia, we observed 5HT-IR; however, contrary to Decapoda, this innervation does not originate from the medulla terminalis, but from a neuropil near the median protocerebrum (Figure 5B). In contrast to Decapoda,  we did not find a division of the HE into two distinct neuropils and an innervation of the medulla terminalis by the OGT. Furthermore, the HE is not organized in distinct layers as in some Decapoda and Stomatopoda [29, 61, 63]. In other Decapoda, the HEs do not show layers but microglomeruli , indicating distinct structural variations of the HEs in Malacostraca. In spite of these structural differences, we suggest that the HEs of Remipedia and Malacostraca are homologous based on their protocerebral position, spherical shape and the connectivity to the deutocerebral ONs by the OGT.
Unpaired midline neuropils of the Protocerebrum
Unpaired midline neuropils are a common feature in the protocerebrum of e.g. Chelicerata, Onychophora [26, 27, 65, 66], Myriapoda , Hexapoda and in crustacean species belonging to Malacostraca [26, 67], Remipedia ([15, 16], this study) and Branchiopoda . In general, the central complex of Tetraconata consists of the unpaired central body, the unpaired protocerebral bridge, and the paired lateral accessory lobes (see neuroanatomical glossary  and ). Utting et al.  performed a detailed study of the central complex of the crayfish and revealed a number of similarities in the overall architecture, neuronal projections and immunoreactivity of neurons, and suggested a possible homology of this complex to that of insects.
Serotonin immunoreactivity in the central complex has been studied in a variety of Hexapoda and Crustacea. In all studied taxa, neuropils of the central complex show 5HT-IR which originates from 5HT-ir neurons in soma clusters situated lateral and posterior to the protocerebral neuropil. In Remipedia, the labeling against serotonin reveals a conspicuous unpaired midline domain sd2 in the center of the protocerebrum (Figure 5E). Furthermore, sd2 is interconnected by fine neurites with an arcuate structure consisting of the paired sd4-6 and by a thick neurite bundle to the laterally positioned sd9. Because of the location and connectivity between these neuropils, which appears similar to that present in other Crustacea and Hexapoda, we consider these structures to be the central complex, comprising the central body (sd2), the protocerebral bridge (sd4-6) and the lateral accessory lobes (sd9). However, we could not find a separation of the central body in an upper and lower division and, similar to all other crustacean taxa, we did not detect noduli as in insects .
In both Crustacea and Hexapoda, 5HT-ir neurons innervate the lateral accessory lobes and then form a commissure near the central body to the contralateral brain hemisphere [67, 70, 71]. Similar neurons are positioned in soma cluster B in Remipedia (sB1; Figure 5B), projecting into the lateral accessory lobes. Whether these neurons contribute to a protocerebral commissure remains unclear. In addition, Remipedia possess various 5HT-ir neurons in soma cluster A, associated with the central complex, especially the central body (Figure 5B). The 5HT-ir neurons sA1 thus may correspond to the central body neurons 1 and 2 (CBN1 and CBN2) in crayfish . Comparable cells are present in Branchiopoda and Insecta [71, 72]. In locusts and crayfish, the protocerebral bridge and the central body are connected via four tracts in each brain hemisphere, the W-, X-, Y-, Z-tracts, containing 5HT-ir neurites [66, 71]. Corresponding neurites have been found in Remipedia (yellow arrowheads in Figure 5D), however, they do not seem to form a chiasm before entering the central body as it is present in locusts. This chiasm is also missing in crayfish . In Remipedia, the tracts connecting protocerebral bridge and central body could be identified only with 5HT-IR. In conclusion, we provide first evidence of W-, X-, Y-, Z-tracts in Remipedia, but further studies are desirable to confirm homology.
The unpaired sd2 in Remipedia receives input from axons of the 5HT-ir neurons sA2 that cross the midline before entering sd2 anteriorly. In Cephalocarida, Malacostraca and Hexapoda corresponding 5HT-ir neurons are situated anteriorly to the central complex. The neurites of these cells decussate before innervating parts of the central body [33, 40, 71]. Unpaired 5HT-ir domains in the center of the protocerebrum have also been described for Cephalocarida and Mystacocarida [30, 33], but other compartments of the central body are missing in these species. Therefore, the described distribution and innervation pattern of the central complex by 5HT-ir neurons seems to be a shared feature in Remipedia, Malacostraca and Hexapoda.
In Arthropoda, the neuromodulator and neurohormone serotonin is involved in a variety of physiological processes, and a circadian fluctuation of serotonin levels has been observed in a number of crustacean taxa [73, 74]. This could be an explanation for the differences in the results of 5HT-IR in Speleonectes tulumensis and Godzilliognomus frondosus. In Decapoda, it was shown that the serotonin level fluctuates in a bimodal or trimodal manner, indicating circadian control [75, 76]. Sandeman et al.  revealed a 5HT-ir connectivity of the protocerebral bridge, the medulla terminalis and the sinus gland in the eyestalk of Cherax destructor. Removal of this gland in Procambarus clarkii caused changes in the circadian rhythm [77, 78]. Remipedia show similar 5HT-ir projections between the protocerebral bridge and the medulla terminalis.