Masters of communication: The brain of the banded cleaner shrimp Stenopus hispidus (Olivier, 1811) with an emphasis on sensory processing areas

The pan‐tropic cleaner shrimp Stenopus hispidus (Crustacea, Stenopodidea) is famous for its specific cleaning behavior in association with client fish and an exclusively monogamous life‐style. Cleaner shrimps feature a broad communicative repertoire, which is considered to depend on superb motor skills and the underlying mechanosensory circuits in combination with sensory organs. Their most prominent head appendages are the two pairs of very long biramous antennules and antennae, which are used both for attracting client fish and for intraspecific communication. Here, we studied the brain anatomy of several specimens of S. hispidus using histological sections, immunohistochemical labeling as well as X‐ray microtomography in combination with 3D reconstructions. Furthermore, we investigated the morphology of antennules and antennae using fluorescence and scanning electron microscopy. Our analyses show that in addition to the complex organization of the multimodal processing centers, especially chemomechanosensory neuropils associated with the antennule and antenna are markedly pronounced when compared to the other neuropils of the central brain. We suggest that in their brains, three topographic maps are present corresponding to the sensory appendages. The brain areas which provide the neuronal substrate for these maps share distinct structural similarities to a unique extent in decapods, such as size and characteristic striated and perpendicular layering. We discuss our findings with respect to the sensory landscape within animal's habitat. In an evolutionary perspective, the cleaner shrimp's brain is an excellent example of how sensory potential and functional demands shape the architecture of primary chemomechanosensory processing areas.

how sensory potential and functional demands shape the architecture of primary chemomechanosensory processing areas.

| INTRODUCTION
The banded cleaner shrimp Stenopus hispidus (Olivier, 1811) is one of the most popular and thus intensely traded species of marine ornamental shrimps with a pan-tropic distribution (Dudoit et al., 2018;Healy & Yaldwyn, 1970). Its other common names include coral banded shrimp, white-banded shrimp, red-banded shrimp, boxer shrimp, bandana shrimp, or barber pole shrimp. This iconic species is famous for its cleaning behavior (Becker, Curtis, & Grutter, 2005;Jonasson, 1987) of different species of coral reef fish, but scientific reports with behavioral observations are scarce (Spotte, 1998) and video recordings are only available from private aquarists and fish-keepers (see e.g., youtube).
The banded cleaner shrimp is highly territorial and features monogamous pair formation (Johnson, 1969). The phylogenetic position of Stenopodidea within Malacostraca is uncertain with regard to the Decapoda, because stenopodid shrimps share characters of euphausiids such as the sperm morphology but also characters (referring the internal and external morphology, larval development and behavior) indicating a closer relationship to reptants (reviewed in Goy, 2010). Two recent phylogenomic analyses (Schwentner et al., 2018;Wolfe et al., 2019) concordantly consider Stenopodidea to be the sister-group to all other Caridea (see phylogram in Figure 1a). The first neuroanatomical description of its central brain was provided by Sandeman and coworkers (Sandeman, Scholtz, & Sandeman, 1993) and information on the visual neuropils and the neuropils of the lateral protocerebrum (lPC) (those parts of the brain situated within the eyestalk) was added by Sullivan and Beltz (2004) (reviewed in Sandeman, Kenning, & Harzsch, 2014) and recently by Wolff, Thoen, Marshall, Sayre, and Strausfeld (2017).
Here, we provide a more detailed description of its brain and primary sensory organs associated with the deuto-and tritocerebrum-the antennules and antennae-in S. hispidus. In this study, a broad spectrum of state-of-the-art neuroanatomical methods was used including scanning electron microscopy (SEM), X-ray micro-computed tomography (μCT), histological preparations using conventional staining methods, and immunohistochemical labeling combined with imaging techniques such as brightfield and fluorescence microscopy as well as confocal laser scanning microscopy (clsm) and 3D reconstructions. and their antennal appendages were used for paraffin histology as well as for fluorescence microscopy followed by scanning electron microscopy (3× ♀; 3× ♂).

| Nomenclature
The neuroanatomical nomenclature used in this manuscript is based on Sandeman, Sandeman, Derby, and Schmidt (1992) and Richter et al. (2010) with modifications adopted from Harzsch and Hansson (2008), , Loesel, Wolf, Kenning, Harzsch, and Sombke (2013), Krieger et al. (2015), Schmidt (2016) and Harzsch and Krieger (2018). For simplification, the neuroanatomical descriptions are limited to only one side of the brain and hold true for all specimens studied if not stated otherwise. The description of brain components is given from anterior to posterior along the neuraxis which is ventro-dorsally flexed almost 90 of the body axis, resulting in an upright position of the brain within the cephalon and the anterior part of the brain situated most dorsally relating to the body axis.
For a consistent terminology, here we suggest avoiding the term "optic neuropils" (Hanström, 1925; as well as "optic lobes" (Kenyon, 1896). Even if the Greek "optikós" and the Latin term "visus" have the identical meaning, nowadays, "optic" in the field of vision research refers to the physically refractive components of the eye for the reception of light. To emphasize the perceptive character of these neuropils, we suggest using the term "visual neuropils" which is consistent with, for example, the visual cortex in mammals, formerly also termed "optic" cortex (Spiller, 1898). All postretinal components that are related to vision, such as the "optic tract" and the "inner" as well as the "outer optic chiasm" should be consequently renamed, too. Here, we suggest to use "visual tract" (VT) and the "inner" (iCh) as well as the "outer visual chiasm" (oCh) accordingly.
The term "esophageal connective" and the corresponding abbreviation OC (British English) are maintained here for simplicity. The syncerebral brain mass excluding the neuropils of lPC and visual neuropils within the eyestalks (see Krieger et al., 2015) is termed "central brain" throughout the text according to Schmidt (2016).
The PT connects the terminal medulla with the median protocerebrum and only a small portion of axons within is represented by the olfactory globular tract (OGT) which connects the deutocerebrum with the lPC. The VT according to Sandeman et al. (1992) and as used here, represents the connection between the visual neuropils and the neuropils of the lPC.

| X-ray microscopy and 3D reconstruction
Two adult specimens of S. hispidus were anesthetized by chilling them at −18 C immersed in sea water for a few minutes and subsequently killed by transferring them into plastic tubes containing Bouin's solution (10% formaldehyde, 5% glacial acetic acid in saturated aqueous picric acid). After 30 min, the antennules, antennae, maxillipeds, pereiopods, and the pleon of each specimen were removed, and finally, the cephalothorax was transferred back into fresh Bouin's fixative for 3 days in a refrigerator (4 C). After fixation, the specimens were gradually dehydrated in an ascending ethanol series at RT (30, 50, 60, 70, 80, 90, 96%, and 3× Sombke, Lipke, Michalik, Uhl, and Harzsch (2015) with slight modifications after Krieger and Spitzner (2020).
The micro-computed tomographic scans were performed with a lab-based X-ray-microscope (Xradia Carl Zeiss Microscopy GmbH,Jena,Germany). The dried samples of S. hispidus were mounted on plastic welding rods using hot glue and were scanned with a 4× (30 kV, 200 μA, 1 s) and a 10× (40 kV, 200 μA, 3 s) objective. For all scans, Binning 2 was applied (summarizing four pixels for noise reduction). Projections obtained by the tomography were reconstructed using the software XMReconstructor (Carl Zeiss Microscopy GmbH, Jena, Germany). To avoid subsequent information loss, Binning 1 (full resolution) was applied for the following reconstruction resulting in image stacks of 993 × 993 pixels and a pixel size of about 4.8 μm (4×) and 1.9 μm (10×), respectively.
Volume reconstruction and visualization was carried out using Amira 5.6.0 and 6.0.1 (FEI Visualization Science Group, Burlington, VT; RRID: SCR_007353) on the basis of μCT-data. Brain structures such as neuropils, nerves, and cell clusters were segmented manually for volumetric analysis and visualization. 3D surfaces corresponding to the segmentation were generated using unconstrained smoothing (Amira: SurfaceGen). Voxel data of the reconstructed neuropils were extracted by using Amira's material statistics tool. The overall appearance of specific brain structures within the cephalon could be visualized based on μCT by using the Amira Volren-module being connected to manually segmented labels and different transparencies.

| Macro-photography
Overview images were taken with a Canon EOS 70D, equipped with a Canon MP-E 65 mm macro objective and Canon Macro Twin Lite MT-24 EX flash. Polarization filters were used in front of light sources and lenses to reduce reflections and enhancing color contrast. The specimens were photographed in ethanol and fixed with a cover slip.
To generate consistently sharp images, several single photographs were taken along the z-axis and subsequently fused with Combine ZM 1.0 (Hadley, 2008). Further image processing was done with Adobe Photoshop CS3 (RRID: SCR_014199).

| Scanning electron microscopy of sensory appendages
For SEM, the antennules and antennae of a total of six specimens of S. hispidus were fixed in 70% ethanol cut into pieces of 20 mm length and dehydrated in a series of graded alcoholic solutions (70-99.5%).
After critical point drying, the samples were mounted on a standard pin stub equipped with adhesive carbon tabs and a small amount of ACHESON silver paint (Cat. G301, G3347, and G3692; Plano GmbH, Marburg, Germany) and examined with a Zeiss EVO LS10 (Carl Zeiss, Oberkochen, Germany) at the Imaging Center of the University of Greifswald. Images and illustrations were produced and processed in Adobe Photoshop (RRID: SCR_014199) and Illustrator CS4 (RRID: SCR_010279) by using general image enhancement features such as black-and-white inversion, brightness, and contrast.

| Immunohistochemistry
Specimens of S. hispidus were anesthetized by chilling them at −18 C immersed in sea water for a few minutes followed by decapitation and dissection of the brain including the brain parts situ- For whole mount preparations coupled with immunolabeling, we used the protocol after Ott (2008) based on a zinc-formaldehyde (ZnFA) fixation. For this method, two adult specimens of S. hispidus were processed by chilling animals being submersed in seawater for several minutes at −20 C until movement stopped. By substituting the seawater with formalin (4% formaldehyde in seawater) the animals were immediately killed and the brains were subsequently dissected in HEPES-buffered saline (HBS) to avoid precipitation of zinc phosphate (which would occur in PBS). The brains were fixed in ZnFA (Cat-No. 15675;Electron Microscopy Sciences, Hatfield, PA) for 20 hr at room temperature on the shaker. After subsequent wash steps for 3 × 15 min in HBS, the brains were postfixed and dehydrated in Dent's fixative (20% dimethyl sulfoxide [Cat-No. 20385; Serva Electrophoresis, Heidelberg, Germany] and 80% methanol) for 2 hr on the shaker followed by transfer into 99% methanol. The samples were then gradually rehydrated in TRIS-buffer with decreasing grades of methanol (90, 70, 50, 30%, and finally pure TRIS-buffer for 15 min each). Samples were preincubated for 4 hr in PBS-TX at room temperature followed by the primary antisera (as listed above) for 4 days at 4 C. After subsequent washing for 4 × 30 min in PBS-TX at RT on the shaker, incubation in the secondary antibodies (as listed above) was carried out for another 2.5 days at 4 C. The brains were then washed for 2 × 1 hr in PBS-TX followed by dehydration in a graded series of ethanol (30, 50, 70, 80, 90, 96, and 2 × 99.5% for 30 min each).

| Antibody reporting
The monoclonal anti-SYNORF1 synapsin antibody (DSHB Hybridoma Product 3C11; anti SYNORF1 as deposited to the DSHB by E. Buchner, University Hospital Würzburg, Germany; supernatant) was raised against a Drosophila melanogaster GST-synapsin fusion protein and recognizes at least four synapsin isoforms (70,74,80,and 143 kDa) in western blots of D. melanogaster head homogenates (Klagges et al., 1996). Sullivan, Benton, Sandeman, and Beltz (2007) mention a single band at~75 kDa in a western blot analysis of crayfish brain homogenate. Harzsch and Hansson (2008) conducted a western blot analysis comparing brain tissue of D. melanogaster and the hermit crab Coenobita clypeatus (Anomala, Coenobitidae). The SYNORF1 serum provided identical results for both species and it stained one strong band between 80 and 90 kDa and a second weaker band slightly above 148 kDa, suggesting that the epitope that SYNORF1 recognizes is strongly conserved between D. melanogaster and C. clypeatus (see Harzsch & Hansson, 2008). Similar to the fruit fly, the antibody consistently labels brain structures in other major subgroups of the malacostracan crustaceans (e.g., Harzsch, Anger, & Dawirs, 1997;Beltz et al., 2003;Meth et al., 2017) in a pattern that is consistent with the assumption that this antibody labels synaptic neuropils in crustaceans. In the following, the term "synapsin-like immunoreactivity" is used to indicate that the antibody most likely recognizes the same isoforms of synapsins as have been shown for the land hermit crab Coenobita clypeatus. However, we used the anti-SYNORF1 synapsin antibody as a morphological marker to label neuropils within the brain of S. hispidus.
It has previously been used to localize A-type allatostatin-like peptides in crustacean and insect nervous systems (e.g., Kreissl, Strasser, & Galizia, 2010;Polanska, Tuchina, Agricola, Hansson, & Harzsch, 2012). In the following, the term "allatostatin-like immunoreactivity" is used to indicate that the antibody most likely binds to various related peptides within this peptide family.
In control experiments for secondary antibody reactivity, in which the primary antibodies were replaced with PBS-TX, no neuronal labeling was detected.

| Histology
For investigating the internal organization of sensory neurons associated with the olfactory sensilla (aesthetascs), we analyzed sagittal Azan-stained paraffin sections of the antennules of three specimens.
After fixation in 70% ethanol, the paraffin embedding procedure involved dehydration of the samples in a series of graded alcohol solutions (70-99.5% ethanol), followed by xylene, and two consecutive infiltrations of paraffin (1 and 2 hr) that was kept at 59-61 C during this process. The samples were then embedded in fresh paraffin. Sections were cut in a sagittal plane at a thickness of 6 μm using a microtome (Leica RM 2145; Leica Microsystems, Wetzlar, Germany), stained with Azan after Geidies (1954), and embedded in Roti-Histokit (Cat. 6638.2; Carl Roth, Karlsruhe, Germany).
For the visualization of somata of olfactory sensory neurons, one antennule was dissected and the lateral and medial flagella were removed distally. The sample was fixed in 70% ethanol for 24 hr at room temperature and was stored at 4 C. The antennule was then rehydrated in decreasing grades of ethanol (50 and 30%) for 30 min each, and finally twice in PBS. It was subsequently incubated with 0.5 μl SYTOX™ Green (Thermo Fisher Scientific-Invitrogen™ Cat. No. S7020) in 1 ml PBS-TX for 24 hr at room temperature and dehydrated again in a graded series of ethanol (30, 50, 70, 80, 90, 96, and 2 × 99.5% for 30 min each).
For silver impregnations of the central brain, the anterior of the cephalothorax was removed from the animals and the exposed brains were fixed in cold, aged alcoholic Bouin's fixative for 2 to 3 days after which they were washed and dissected free from the surrounding tissues. They were dehydrated in an ethanol series, cleared in xylene and embedded in paraffin wax. 10 μm serial sections were mounted on glass slides and stained using a modification of the Holmes-Blest silver impregnation method (Blest & Davie, 1980), in which impregnation times were increased up to 24 hr and toned with a 2% gold solution. For further details, see Sandeman et al. (1993).

| Microscopy and confocal laser scanning
Brightfield micrographs of histological preparations (paraffin sections stained with Azan and silver impregnated paraffin sections) as well as epifluorescence micrographs of the antennules and antennae of S. hispidus were performed using a Nikon Eclipse 90i fluorescence microscope equipped with two mounted digital cameras for black-and-white (Nikon DS2-MBWc) and color photos (Nikon DS-Fi3) that were generated by the use of the software package NIS-Elements AR 5.02.00 (RRID: SCR_014329). Cuticular auto-fluorescence of the antennules and antennae was excited with ultraviolet light (UV) with a wavelength of 340-380 nm eliciting light emissions with a wavelength of 435-485 nm.
To minimize the usage of animals, fluorescence micrographs of the antennules and antennae were followed by the sample preparation for scanning electron microscopy (see above) using the same samples. scaphocerite is almost as long as the carapace. Although we recorded the sex of most individuals, no apparent sexual dimorphism within the brain or sensory appendages could be revealed based on our analyses, and further insights into this topic are beyond the scope of this manuscript.

| Sensory appendages associated with the brain
With about 120 mm length, the antennule (A1) as well as the antenna (A2), both extend anteriorly over more than twice the body-length of the animal (60 mm) and can sample across almost the entire dorsal hemisphere of the animal. The antennule (Figure 1a ± 11 (average ± SD; ranging from 70 to 96; n = 6; 81 ± 13 in 3× ♀ and F I G U R E 2 Antennule of Stenopus hispidus. The proximalmost part of the antennule showing the peduncle and the lateral flagellum (lFl) bearing the aesthetascs (as) on its medial side and several armoring spines (white and black arrowheads) on its lateral side (a-c, g) as well as the medial flagellum (mFl) revealed by a scanning electron micrograph of a male specimen in (a). In image section (b), a bright-field micrograph shows the aesthetascs (as) and cuspidate setae (cs) in higher magnification on annuli 14 to 18 on the lFl of an unsexed specimen. A black-white inverted maximum projection of a confocally scanned lFl reveals the ellipsoidal clusters of olfactory sensory neurons (OSNs) in addition with integument cells (flattened cells on top and bottom) labeled with the nuclear marker Sytox™ Green (green) in (c) and in higher magnification in (d) of an unsexed specimen. The scanning electron micrograph of a female specimen in (e) displays the aesthetascs flanked of three hair-like sensilla (asterisks) on the anterior margin of each annulus. The bright-field micrograph in (f) displays a longitudinal paraffin section of a female specimen stained with Azan through the lFl revealing the aesthetascs (orange) as well as the ellipsoidal clusters of OSNs (purple) beneath the cuticle (blue) corresponding to (c) and (d). The scanning electron micrograph in (g) shows several armoring spines on lateral side of lFl in greater detail and a few cuspidate setae (cs) [Color figure can be viewed at wileyonlinelibrary.com] 86 ± 9 in 3× ♂) unimodal chemosensilla (Figures 2a-c,e and 3a-d; the so-called aesthetascs are identifiable but are absent on the median flagellum contradicting the description by Goy (2010). The aesthetascs (as) which mediate olfaction in crustaceans (Derby, Kozma, Senatore, & Schmidt, 2016) are arranged in parallel rows of two to three aesthetascs per row (the number increases gradually from distal to proximal thus ranging from two to six rows per annulus) at the medioventral side of the lFl in S. hispidus. However, several pores of F I G U R E 3 Scanning electron micrographs of the antennular sensilla. SEM-micrographs of a lateral flagellum (lFl) of a male specimen show numerous pores at the insertion region of aesthetascs (as) in (a) and in higher magnification (dotted rectangle in a) is shown in (b), whereas no apical pores on the aesthetascs (as) could be identified as shown from a lFl in a female specimen in (c) and in higher magnification (dotted rectangle in c) in (d). Note that the lFl in (c) and (d) is orientated upside-down (aesthetascs face upwards) in contrast to the other panels. Fluorescence micrographs using the cuticular autofluorescence display the arrangement and distribution of cuspidate setae (cs) for the middle region each of the lFl in (e); the median flagellum (mFl) in (f); and the antennal flagellum (Fl) in (g) of an unsexed individual of S. hispidus. SEMmicrographs of the antennular lFl (in h) and mFl (in i) of a male specimen display each a cuspidate seta (cs) in addition of a simple seta (ss) in proximity to cuticular pores of unknown function unknown function occur in a crescent-shaped arrangement at the cuticular bulge of the lFl under which the aesthetascs originate (shared socket; see Figure 3a,b). These aesthetasc rows are occasionally flanked by very slender hair-like sensilla (simple setae according to Garm & Watling, 2013)  Each cuspidate seta is articulated in small depressions or pits in the cuticle (infracuticular articulation), has a compact appearance (about 15 μm diameter at the setal base), tapers distally, and is about 50 μm long. In addition, each cuspidate seta is associated with one long and slender setule (sensory bristle according to Brandt, 1988) branching off on the distal half of the seta and exceeding its length by about 50 μm (see Figure 3h,i). The robust cuspidate seta (L:W ratio of about 4.5) does not possess a terminal pore. Based on its outer morphology, the cuspidate seta resembles those of the "sensory spines" described as mechanosensory sensilla in isopods such as Aega antarctica (Brandt, 1988   F I G U R E 5 Visualization of the brain within the head of a male specimen of S. hispidus. The general outline of the 3D reconstructed brain is illustrated based on manually segmented virtual sections based on μCT using volume-rendering "Volren" combined with "Volume rendering" (Amira 6.0.1) from dorsal in (a) and anterolateral in (b). In a posterior view in panel (c), the posterior half of the head is cut away (using "Orthoslice"). Panel (d) shows a lateral view on the same specimen while one half of the head is sagittally cut away. Abbreviations: A I Nv, antennular nerve; A II Nv, antennal nerve; AnN, antennal neuropil; LAN, lateral antennular neuropil; lPC, lateral protocerebrum; OC, esophageal connective; OL, olfactory lobe; VNs, visual neuropils [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 6 Overview of the visual neuropils (VNs) and the neuropils of the lateral protocerebrum (lPC) and central brain in S. hispidus. The general structuring of the cleaner shrimp's brain is depicted in a schematic drawing in (a) Micrographs of specific regions (dotted rectangles in a) are shown in the panels (b to d). Panel (b) shows the lateral protocerebral neuropils (note that the lamina was destroyed in this preparation due to dissection) of a male specimen in a frontal vibratome section labeled against anti-SYNORF1 synapsin (SYN; black-white inverted fluorescence micrograph). Two opposed hemispheres of the central brain are displayed equally-scaled based on a silver-impregnated paraffin section of an unsexed specimen (c) and an almost frontal vibratome-section of a female specimen triple labeled against anti-SYNORF1 synapsin (SYN; magenta) and anti-allatostatin (AST; green) in addition with the nuclear counterstain HOECHST (NUC; blue) in ( F I G U R E 7 The visual neuropils and the neuropils of the lateral protocerebrum in S. hispidus. 3D reconstruction of the neuropils of the visual neuropils and the neuropils of the lateral protocerebrum based on manual segmentation (surfaces) in virtual slices based on μCT of the eyestalk are shown from posterior (a) and anterior (b) of an unsexed specimen; in addition of a frontal vibratome-section triple labeled against anti-SYNORF1 synapsin (SYN; magenta) and anti-allatostatin (AST; green) and the nuclear counterstain HOECHST (NUC; blue) of hemiellipsoid body (HN) and terminal medulla (TM) in a slightly sagittal orientation of a female specimen (rectangle in a) in (c) showing the "lateral protocerebral complex" (arrowheads in a to c) after Sullivan and Beltz (2004). A virtual section of the same μCT of an unsexed specimen is displayed in (d) and opposed to a frontal vibratome-section (using the same labeling as in c) of the hemiellipsoid body (HN) and terminal medulla (TM) of a female specimen in (e) (note that the lamina was destroyed in this preparation due to dissection) and of higher magnifications (dotted rectangles in e and f) in (f) and (g). Other abbreviations: asterisk, central region of the HN sublobes; 5; cell cluster (5) become visible by X-ray microscopic tomograms (asterisks in Figure 7d,f) presumably originating in the somata cluster (5) displaying weak or no ASTir too. At the periphery, in higher magnifications, a network of synaptic profiles with regular swellings resembling microglomerular structures (2 to 3 μm in diameter) becomes identifiable by SYNir which is colocalized by ASTir (arrows in Figure 7g) in three clear strata (Figure 7e) of which the centralmost shows the weakest allatostatinergic signal intensity (Figure 7g). At the periphery of the cortex, allatostatinergic neurites appear to interconnect the microglomeruli peripherally (Figure 7g). Immunohistochemically labeled sections revealed that cluster (5) is almost devoid of ASTir (Figure 7c,e).
Cell cluster (5) which houses densely packed somata of small globuli F I G U R E 9 Legend on next page. cells (a term introduced by Holmgren (1916) for interneurons in polychaetes, onychophorans, and euarthropods in mushroom-body-like structures and now known as Kenyon cells in insect mushroom bodies, see Strausfeld (1976)) This cluster (5) covers anteriorly more than a half of the HN A and further ventrally also parts of the large terminal medulla, where few somata can be identified showing distinct ASTir.
The medioventral compact terminal medulla shows strong but diffuse SYNir, a network of strong ASTir and its entire volume slightly exceeds the volume of all three hemiellipsoid neuropils together (Table 1) which Sullivan and Beltz named "lateral protocerebral complex" (Sullivan & Beltz, 2004). From a medio-dorsal perspective, this neuropil has a knob-like shape at the margin of the terminal medulla and extends into the hemiellipsoid body neuropils forming a strong curvature.

| Median protocerebrum (mPC)
Each lPC is connected medioventrally to the ipsilateral portion of the

| Deutocerebrum
The deutocerebrum, which is associated with the antennule (synonym for antenna 1), is the most partitioned brain region in S. hispidus. It  (Figures 9a and 10a,b), all four neurite bundles can be tracked F I G U R E 9 The olfactory lobe (OL) in S. hispidus. A 3D volume rendering ("Volren") of the central brain of an unsexed specimen based on manual segmentation in virtual slices based on μCT is shown in (a) highlighting the lateral deutocerebrum and antennal neuropils (AnN-green). (b) shows a manually segmented 3D-reconstruction of the olfactory glomeruli (og) in the OL of a female specimen based on a whole-mount preparation of the brain of S. hispidus double-labeled against anti-SYNORF1 synapsin (colored surfaces) and the nuclear counterstain HOECHST (white). (c) displays a frontal silver-impregnated paraffin-section of the OL and its olfactory glomeruli (dotted rectangle in c is shown in d at a higher magnification). (e) displays a maximum projection of a confocal image stack of the OL based on a vibratome-section triple labeled against anti-SYNORF1 synapsin (SYN; magenta) and anti-allatostatin (AST; green) and the nuclear counterstain HOECHST (NUC; blue) revealing allatostatinergic neurites of projection neurons (black arrowhead) exiting the medial foramen (mf) via the olfactory globular tract in a male specimen. The dotted rectangle in (e) is displayed at a higher magnification in (f) (hiding the green channel) showing the substructure of og into a base, subcap, and cap region. Panels (g) (same orientation as in e) and (h) (orientation according to the dotted parallelogram in e) show interglomerular contacts (arrowheads) in virtual sections from confocal scans of two more preparations of a female specimen using the same labeling as shown in (e). Other abbreviations: 6, 10, 9/11; cell clusters (6), (10), (9/11); A I Nv, antennular nerve; A II Nv, antennal nerve; lLANNv and mLANNv, lateral and medial nerves innervating the lLAN and mLAN, lateral and medial lobe of the lateral antennular neuropil; lf, lateral foramen; lPC, lateral protocerebrum; MAN, median antennular neuropil; MANNv, median antennular nerve; mPC, median protocerebrum; OC, esophageal connective; OLNv, aesthetasc nerve innervating the OL. All scale bars represent 50 μm as shown in (b) [Color figure can be viewed at wileyonlinelibrary.com] to their specific target neuropils which are from lateral to median for each hemisphere: one nerve innervating the median antennular (antenna I-) neuropil (MANNv), one nerve the olfactory lobe (OLNv), and two nerves (lLANNv and mLANNv) innervating the lateral (lLAN) and the medial lobes (mLAN) of the lateral antennular neuropil (LAN).
The unpaired median antennular neuropil is confluently interconnected between the posterior medial protocerebral neuropil of the median protocerebrum anterodorsally, the lateral antennular neuropil laterally, and the tritocerebrum posteroventrally, resembling an inverted "U" in frontal sections (Figure 10c-d1). The transition between posterior medial protocerebral neuropil and median antennular neuropil can be identified by the presence of the cerebral artery penetrating the central brain in an anterodorsal course. In some decapods, the median antenna antennular neuropil receives input from the statocyst on the basal antennular peduncle (which is lacking in Stenopodidea [Goy, 2010]) and other mechanosensory sensilla of the F I G U R E 1 0 Legend on next page. F I G U R E 1 0 Deuto-and tritocerebral neuropils processing chemomechanosensation in S. hispidus. A 3D reconstruction as volume rendering ("Volren") of the central brain based on manual segmentation in virtual slices based on μCT of a male specimen from ventral (see arrowhead in pictogram) is shown in (a), a μCT-based visualization of the central brain of an unsexed individual at higher magnification is shown in (b) and illustrated as surfaces from posterior in (c) (see arrowhead in pictogram). (d) shows a frontal (see pictogram) silver-impregnated paraffin section through the central brain of an unsexed specimen revealing the median deutocerebrum (dotted rectangle in d at higher magnification in d1). In a silver-impregnated paraffin section further anterior (see pictogram) in (e), the antennal neuropil (AnN) of the tritocerebrum (TC) is revealed. (f) displays a slightly horizontal vibratome section (see pictogram) of the antennular and antennal chemomechanosensory processing centers triple labeled against anti-SYNORF1 synapsin (SYN; magenta) and anti-allatostatin (AST; green in f and black in a black-white inverted single channel in f1) and the nuclear counterstain HOECHST (NUC; blue) and the lateral (lLAN) and medial lobe (mLAN) of the lateral antennular neuropil (LAN) of a female specimen and of another female specimen at a higher magnification in (g). Other abbreviations: 6, 10; cell clusters (6)  Note: Volumetric data (based on μCT of three specimens; one male; one female and one unsexed specimen) as well as the number of ommatidia (based on a male specimen) were generated based on a 3D reconstruction of a μCT. The relative volumes of neuronal structures refer to both hemispheres whereas the estimated numbers of sensory organs and corresponding sensilla refer to only one hemisphere. *The number of retinula cells results from the ommatidia count multiplied by eight retinula cells according to the tetraconate ground pattern after Melzer, Diersch, Nicastro, and Smola (1997). **The number of olfactory glomeruli (og) results from the volume of one olfactory lobe (OL; SYNir signal of a confocal scan of one whole-mount preparation) divided by the average volume of 20 randomly chosen og therein. The number of aesthetascs is an average of a total of six individuals (three males and three females; of one lateral flagellum per individual only) manually counted based on scanning electron micrographs. The number of cuspidate setae is estimated based on the number of annuli (counted in one individual were all three flagella were intact) multiplied by the average number of cuspidate setae per annulus (based the number of cuspidate setae counted on nine annuli in different flagellar regions [apically, medially, and basally]). Abbreviations: lFl, lateral antennular flagellum; mFl, medial antennular flagellum; Fl, antennal flagellum. antennular base . Furthermore, antennular motoneurons project from the median antenna antennular neuropil into the antennule . In S. hispidus, the median antennular neuropil shows distinct SYNir and weak ASTir (Figure 6d).
Neurites with strong ASTir traverse it dorsoventrally. Anterolateral to median antennular neuropil, the lateral antennular (antenna I-) neuropil originates and projects anteriorly (Figure 9a). Ventral between the olfactory lobe and the lateral antennular neuropil, a cluster (16) (Figure 6c). The minute elliptical OGT neuropil is situated in close proximity of the OGT between the median foramen and lateral antennular neuropil. An accessory lobe (AcN) as present as a prominent structure in Astacidea, or rather minute compared to the size of the OL in Achelata, Axiidea, Anomala, and Brachyura (Sandeman et al., 1993) could not be detected in any of our preparations.

| General
The brain organization ( Figure 4) and its innervation from the primary sensory centers in Stenopus hispidus comprise all elements of the suggested malacostracan ground pattern Sandeman et al., 2014). The schematic illustrations showing the brain of S. hispidus given in Sandeman et al. (2014), however, underestimated the size of the lPC compared to the central brain, perhaps because a previous study referring to Stenopus among other malacostracans has focused on the central brain only (Sandeman et al., 1993). The volume of both lateral protocerebra taken together are larger than the central brain in the cleaner shrimp. The multimodal association centers are composed of the hemiellipsoid body and the terminal medulla dominate the brain (Figure 4 and Table 1). The gross brain anatomy is comparable to that of the dendrobranchiate Pacific White Shrimp Penaeus vannamei Boone, 1931, as described by Meth et al. (2017) although specific neuropils differ considerably in shape and size. In the Pacific White Shrimp, the lamina is thinner but stretches like a cup over a larger area than in the banded cleaner shrimp whereas all other visual neuropils such as the medulla, lobula, and lobula plate are of comparable sizes in both species. The lPC is more pronounced in S. hispidus and its separation into a terminal medulla and a hemiellipsoid body with complex substructures are absent in P. vannamei (Meth et al., 2017). Hence, the organization of the stenopodean hemiellipsoid body into columnar and calycal neuropils including microglomeruli resembles that of representatives of Caridea (Sayre & Strausfeld, 2019) more closely than that of the Dendrobranchiata (Meth et al., 2017;Sullivan & Beltz, 2004).  et al., 2017) and in our view, reflects a size relationship between primary sensory structures and their associated neuropils. Consequently, in P. vannamei, the antenna (synonym for antenna 2) and its tritocerebral processing unit, the antennal neuropil are comparably large so that it dominates the central brain (Meth et al., 2017).
In summary, the neural architecture of S. hispidus supports its phylogenetic position according to recent phylogenomic analyses Figure 1a (Schwentner et al., 2018;Wolfe et al., 2019), since it shares more similarities with the caridean brain than with its dendrobranchiate relatives.

| The hemiellipsoid body and possible genealogical relations to insect higher order neuropils
Decapod crustaceans display a rich repertoire of complex behavioral patterns related to finding food, shelter, and mating partners, kin recognition and brood care, orientation and homing, and are also known for complex social interactions such as communal defensive tactics, the occupation of common shelters, cooperative behavior during seasonal long-distance migration and the establishment of dominance hierarchies (reviews Duffy & Thiel, 2007;Breithaupt & Thiel, 2011;Derby & Thiel, 2014;Thiel & Watling, 2015). Higher order integrative brain centers are suggested to play a major role in learning and memory as well as to provide the neuronal substrate for more sophisticated processing underlying such behaviors (review in Sandeman et al., 2014). These authors suggested that in particular, those behaviors that require an active exploratory component and involve 3D spatial perception (be it visual, tactile, or olfactory), require the involvement of more than one of such higher order brain centers. Higher order neuropils receive input mostly from second or third order neurons but are not targeted by any primary sensory afferents and contain interneurons that respond to the stimulation of several different sensory systems. In the malacostracan brain, the (bilaterally paired) terminal medulla, hemiellipsoid body, accessory lobe, and the unpaired central complex seem to function as higher integrative centers, all four distinct neuropil areas which display a high level of complexity and are notable for their substantial volume (Sandeman et al., 2014).
The lPC and in particular, the hemiellipsoid body of S. hispidus ( Figure 7) is exceptional in comparison to all other malacostracans studied as has already been noted by Sullivan and Beltz (2004). The size of the lPC is enormous in relation to the central brain ( Figure 4 and Table 1) and also in comparison to other malacostracans studied. The total neuropil volume of the terminal medulla and hemiellipsoid body together (of both hemispheres) represented 45.9 ± 1.5%, and together with the visual neuropils 59.5 ± 2.4% of the total brain volume (average ± SD; n = 3; Table 1). All three sublobes are characterized by peripheral microglomerular synaptic networks, clearly regionalized into an outer dense and inner loose cortex of synaptic microglomeruli, interspersed with peripheral ASTir profiles in different intensities resulting in a highly ordered pattern (Figure 7d-g). This feature of the cortices resembles the organization of the calyx region of several hexapod mushroom bodies (Groh & Rössler, 2011), higher order multimodal neuropils in the hexapod protocerebrum (Fahrbach, 2006;Farris, 2005;Galizia, 2008;Martin et al., 2011;Strausfeld, 1998;Strausfeld, Sinakevitch, Brown, & Farris, 2009). Such similarities include as a regional distribution of different microglomerular densities and an average size of microglomeruli between 2 and 3 μm in diameter in by Sullivan and Beltz (2004). Based on its position, shape, and pattern of immunohistochemical labeling, the lateral protocerebral complex resembles that of the insect mushroom bodies' peduncle; in addition, its knob-like terminal (arrowheads in Figure 7a-c) resembles that of the ventral lobe like in the stick insect Sypiloidea sypilus (Westwood,
Discussing possible genealogical relationships of insect mushroom bodies and the crustacean hemiellipsoid body/terminal medullacomplex has entertained generations of arthropod neuroanatomists for almost 100 years (see references in Strausfeld, Hansen, Li, Gomez, & Ito, 1998). Whereas Strausfeld and coworkers  and Harzsch (2006) were hesitant to assign a homology of these protocerebral centers, a study on the hemiellipsoid body of the hermit crab Coenobita clypeatus identified possible similarities of the neuronal core circuits in this neuropil with those in the insect calyx and peduncle (Wolff, Harzsch, Hansson, Brown, & Strausfeld, 2012). More recently, a reanalysis of the brain in Remipedia, the potential sister group to Hexapoda, also suggested structural similarities of their hemiellipsoid bodies and insect mushroom bodies (Stemme, Iliffe, & Bicker, 2016). In addition, Maza and coworkers (2016) proposed such a genealogical correspondence between insect mushroom bodies and the hemiellipsoid bodies in the brachyuran Neohelice granulata. Along the same lines, Wolff et al. (2017) considered the "lateral protocerebral complex" of S. hispidus a modified version of the ancestral columnar mushroom body of insects because the knob-like terminal of the lateral protocerebral complex displays DC0-immunoreactivity. However, these authors interpreted their results with caution and thoroughly discussed a possible convergent evolution of the structural similarities of insect mushroom bodies and the crustacean hemiellipsoid body/terminal medulla-complex (Wolff et al., 2017). Any discussion on this topic must account for the phenotypic diversity which both insect mushroom bodies (Farris, 2005;Strausfeld et al., 2009) (Sayre & Strausfeld, 2019)

and the hermit crab
Pagurus hirsutiusculus (Strausfeld & Sayre, 2020). Again, common Leitmotifs were identified in the structural elements of these protocerebral centers. We conclude that, although the lPC shows a high variability within Malacostraca (reviewed in Machon et al., 2019), the anatomical similarities we describe here, its position, as well as its interconnection in the primary olfactory pathway suggests that insect mushroom bodies and crustacean hemiellipsoid bodies share neural circuit motifs.
These include the presence of numerous local interneurons (intrinsic neurons) the neurites of which interact with afferences provided by one or several distinct tracts of projection neurons from the primary olfactory neuropils in the deutocerebrum. Despite these shared core circuits, the phenotype of these complex protocerebral centers nevertheless diverged greatly during the evolutionary diversification of the hexapod and malacostracan lineages so that structural similarities are difficult to unravel.
Functionally, the malacostracan lPC is supposed to be an association center because it receives multimodal input of higher-order (no primary afferents). Its connections with the olfactory pathway were described in numerous anatomical and physiological studies (e.g., Derby & Blaustein, 1988;Mellon, Sandeman, & Sandeman, 1992;Sullivan & Beltz, 2001, 2004. Besides its undisputed role as center of a higher-order processing, a recent study in brachyuran crabs by Maza and coworkers (Maza et al., 2016), suggests that the hemiellipsoid body and terminal medulla are likely to be involved in processes of associative memory formation.
However, concerning the elements of the primary olfactory pathway in the banded cleaner shrimp, it has to be noted that the number of olfactory sensory neurons and aesthetascs as well as the size of olfactory lobes is moderate in comparison to other malacostracans (Schmidt, 2016). Along these lines of argument we suggest that olfactory learning may play a subordinate role in this structure but instead other functional roles must explain the remarkable elaboration of the neuropils of the lPC in this species.
The banded cleaner shrimp shows extraordinary communicative skills (intra-as well as interspecifically) and is suggested to feature aspects of social behavior, for example, by pair formation (Becker et al., 2005;Johnson, 1969Johnson, , 1977Jonasson, 1987;Wolff et al., 2017). Cleaner shrimps have to master identifying a variety of host species and in particular, discriminating interspecific behaviors such as the willingness of a host to be cleaned, even if it is a potential predator. These skills suggest sophisticated learning capabilities with regard to communication which might be reflected in a very pronounced hemiellipsoid body. However, since it is very territorial (Chockley & Mary, 2003), S. hispidus is not considered to display sophisticated navigational skills which are often but not exclusively linked to the need of spatial learning and thus effective multimodal association centers like the mushroom bodies in honey bees or ants (Hexapoda) (Grob, Fleischmann, Grübel, Wehner, & Rössler, 2017;Menzel & Müller, 1996;Webb & Wystrach, 2016), or the hemiellipsoid bodies in mantis shrimps (Stomatopoda) (Wolff et al., 2017), spiny lobsters (Achelata) (Boles & Lohmann, 2003;Steullet et al., 2002), or robber crabs (Anomala) (Krieger et al., 2010. A highly territorial lifestyle, as reported for the banded cleaner shrimp, is a behavior essentially dependent on the realization of good place memory, so that spatial learning could play a major role for these animals. Wolff and coworkers (2017) observed that insects with elaborate navigational skills display elaborate mushroom bodies and suggested that place memory is likely to be processed in the crustacean hemiellipsoid body as well as in the insect mushroom body. This hypothesis fits well with the expansion of the hemiellipsoid body in crustaceans such as the coconut crab Birgus latro (Krieger et al., 2010) for which a seminomadic behavior characterized by territorial phases coupled with phases involving elaborate navigational skills have been reported (Krieger, Grandy, et al., 2012). A similar correlation between comparably large hemiellipsoid bodies (original pictures from M. Schmidt; modified from Schmidt & Ache, 1996; combined and equally scaled in Harzsch & Krieger, 2018) and lifestyles involving elaborate spatial memory (review in Sandeman et al., 2014) can be assigned also to the seminomadic spiny lobster Panulirus argus. Deep hydrothermal vent shrimps Rimicaris exoculata live in an extreme, lightless habitat characterized by steep temperature gradients. A recent study showed that these animals also feature sophisticated hemiellipsoid bodies (Machon et al., 2019), so that these authors suggested that an excellent place memory may be essential for avoiding the dangerously hot vent chimneys and memorizing emission sites of hydrothermal fluids rich in those chemicals on which their endosymbiont bacteria depend.

| Structural and numerical similarities of the primary deuto-and tritocerebral chemomechanosensory pathways
Due to its specific behavior in advertising and cleaning different fish species, the banded cleaner shrimp requires extraordinary communicative skills. In addition to lateral body swaying, S. hispidus is reported to perform vigorous whipping movements with both its antennules and antennae in presence of potential client fish (Becker et al., 2005) suggesting that at least for luring client fish, the very long flagella of antennules and antennae play a major role. Once attracted, the cleaner shrimp keeps contact with its host fish mainly using the walking legs and claws but only occasionally with its antennules or antennae. However, even during cleaning and feeding, the animals constantly probe their dorsal hemisphere by all three pairs of flagella like six flexible white canes. Interestingly, we found a striking correlation (R 2 > 0.99) between the estimated number of cuspidate setae per flagellum (lFl, mFl, and Fl) and the volumes of their corresponding first order processing units in the central brain (lLAN, mLAN, and AnN) which are known chemomechanosensory centers. Within these centers, mechanosensory afferents as well as antennal (antennular) motoneurons terminate at specific discs resulting in a typical striated pattern Sandeman et al., 2014;Schmidt, 2007Schmidt, , 2016Tautz & Müller-Tautz, 1983) as could be recovered in our preparations (Figures 6c-d, 10e-g). This pattern is suggested to reflect a somatotopic arrangement of bimodal sensilla on antennules and antennae (Sandeman et al., 2014). In addition to the mammalian neocortex (Kaas, 1997;Penfield & Boldrey, 1937), topographic representations of different sensory modalities in the central nervous system can be found, for example, in the visual (Heinze & Homberg, 2007), as well as in the gustatory and mechanosensory system (Loesel et al., 2013;Nishino, Nishikawa, Yokohari, & Mizunami, 2005) of hexapods, and in the chemo-and mechanosensory pathway of the pectine organs in scorpions (Drozd, 2019). These examples show that somatotopic maps are a conserved principle for neuronal circuits in a variety of organisms. Nevertheless, for neuropils processing identical sensory modalities, it is not particularly surprising that more peripheral input needs more computational and hence neuronal substrate in the central brain. The positive correlation between the number of ommatidia and volume of optic lobes (visual neuropils) in males and females (Rein, Zöckler, & Heisenberg, 1999)  Achelata. Furthermore, branches of antennular motoneurons were identified in Brachyura  and references therein). However, in S. hispidus which is lacking a statocyst on the antennular peduncle (Goy, 2010), the well identifiable median antennular neuropil must process either antennular nonstatocyst input and/or is an additional center of antennular motor control besides the LAN.

ACKNOWLEDGMENTS
We cordially thank Caroline Viertel for the animal husbandry in the laboratory. We express our gratitude to Erika Becker for performing the paraffin histology and to Matthes Kenning for providing the SEMimage of the antennule of one specimen of S. hispidus. Images of several animals in Figure 1 were kindly provided by Ann-Christin Rath and Rebecca Meth. We gratefully acknowledge Martin Winter for providing preliminary data from an internship as well as Andy Sombke for his supervision during that time. We express our gratitude to two anonymous referees for their fast reviews and constructive criticism.
We furthermore want to thank all people indirectly supporting our