Adult neurogenesis in crayfish: Origin, expansion, and migration of neural progenitor lineages in a pseudostratified neuroepithelium

Two decades after the discovery of adult‐born neurons in the brains of decapod crustaceans, the deutocerebral proliferative system (DPS) producing these neural lineages has become a model of adult neurogenesis in invertebrates. Studies on crayfish have provided substantial insights into the anatomy, cellular dynamics, and regulation of the DPS. Contrary to traditional thinking, recent evidence suggests that the neurogenic niche in the crayfish DPS lacks self‐renewing stem cells, its cell pool being instead sustained via integration of hemocytes generated by the innate immune system. Here, we investigated the origin, division and migration patterns of the adult‐born neural progenitor (NP) lineages in detail. We show that the niche cell pool is not only replenished by hemocyte integration but also by limited numbers of symmetric cell divisions with some characteristics reminiscent of interkinetic nuclear migration. Once specified in the niche, first generation NPs act as transit‐amplifying intermediate NPs that eventually exit and produce multicellular clones as they move along migratory streams toward target brain areas. Different clones may migrate simultaneously in the streams but occupy separate tracks and show spatio‐temporally flexible division patterns. Based on this, we propose an extended DPS model that emphasizes structural similarities to pseudostratified neuroepithelia in other arthropods and vertebrates. This model includes hemocyte integration and intrinsic cell proliferation to synergistically counteract niche cell pool depletion during the animal's lifespan. Further, we discuss parallels to recent findings on mammalian adult neurogenesis, as both systems seem to exhibit a similar decoupling of proliferative replenishment divisions and consuming neurogenic divisions.


| INTRODUCTION
Throughout most of the 20th century, the accepted view was that new neurons are generated only during embryonic and perinatal life in the vast majority of vertebrate and invertebrate organisms. Only after the introduction of S-phase specific in vivo markers that facilitated the tracking of cell proliferation and substantial improvements of immunohistochemical techniques, this dogma was finally discarded (see Gage & Temple, 2013) as functional integration of new neurons was demonstrated in adult brains of a variety of vertebrate taxa (e.g., Grandel & Brand, 2013), including humans (Eriksson et al., 1998). For a long time, however, elucidation of the lineage relationships among the neural progenitor cells (NPs) producing the adult-born neurons proved challenging (e.g., Zhao, Deng, & Gage, 2008), until recently, when first live-imaging approaches in mice have begun to provide a more comprehensive understanding (e.g., Pilz et al., 2018). In parallel with vertebrates, studies on invertebrates (predominantly arthropods) have assessed the presence of adult neurogenesis and investigated the underlying mechanisms (e.g., Simoes, & Rhiner, 2017).
At the level of single adult NP lineages, the current DPS model proposes three spatially separated cell generations Zhang et al., 2009). The first generation NPs in the niche do not self-renew in stem cell-like manner  but data suggested that they divide only once in a morphologically symmetric fashion before both of their daughter cells (= second generation NPs) start to migrate along the streams. The second generation NPs act as transit-amplifying cells, with some of these undergoing mitosis in the streams. The majority then enters M phase in the proliferation zones, where they subsequently divide at least once (= third generation) or potentially multiple times prior to cell cycle exit and neuronal differentiation (e.g., Zhang et al., 2009). One of the most intriguing aspects of this model is the lack of self-renewal in the first generation NPs. This contrasts not only with the embryonic NP lineages in crayfish, which feature asymmetrically dividing, self-renewing neural stem cells (neuroblasts) at their base (e.g., Scholtz, 1992;Sintoni et al., 2012;Sullivan & Macmillan, 2001). It also indicates replenishment of niche cells by an external source, as niche cell number increases with animal size despite continuous NP efflux and lack of additional niche divisions (Zhang et al., 2009). Notably, the DPS shows considerable vascularization (Chaves da Silva, Sullivan et al., 2007) and various lines of evidence support attachment and integration of hemocytes into the niche  as well as their direct involvement in the production of new adult NP lineages .
Given the current lack of a live-imaging approach for the crayfish DPS, division and migration patterns of NP lineages and hemocyteniche interactions cannot be observed directly but need to be reconstructed from fixed brain preparations. To some extent, such reconstructions are made challenging by the more flexible nature of adult neurogenesis compared to embryonic neurogenesis. Proliferation rates in the DPS decrease gradually with animal size (Zhang et al., 2009) but are also susceptible to a variety of other factors. For instance, NP proliferation and/or survival of the adult-born neurons have been shown to be regulated by endogenous levels of serotonin (Benton, Goergen, Rogan, & Beltz, 2008;Sandeman, Benton, & Beltz, 2009;, nitric oxide , and cytokines  and respond to environmental enrichment (Ayub, Benton, Zhang, & Beltz, 2011;Sandeman & Sandeman, 2000) as well as physical activity (Kim, Sandeman, Benton, & Beltz, 2014). This contrasts starkly with embryonic neurogenesis in crayfish and other malacostracan crustaceans, where the basic nervous system components are laid down by segmental NP lineages in a largely stereotyped division sequence (e.g., Ungerer & Scholtz, 2008).
F I G U R E 1 Procambarus clarkii and anatomy of its adult deutocerebral proliferative system (DPS). (a) Live male specimen of P. clarkii. (b) Ventral view of a desheated deutocerebral hemisphere, immunolabeling of acetylated tubulin (white) with nuclear counterstain (blue), maximum projection. The yellow stippled oval outlines the bilobed niche on the ventral surface of the accessory lobe (acl). Yellow arrows indicate the migratory streams toward neuron Clusters 9 and 10 (cl9 and cl10, respectively; cl9 highlighted by white stippled line). Note that the stream toward cl9 curves around the root of the antenna 1 nerve (a1n, black stippled oval). (c) Optical sections through a niche, phalloidin labeling (green) with nuclear counterstain (blue), Imaris section mode. Note strong F-actin signal in the apical tips of Type 1 niche cell processes (arrows) and the uncompressed cell-free cavity lumen as seen in the sagittal and cross sections shown at the right and bottom, respectively. (d) Optical sections through a niche, acetylated tubulin immunolabeling (white) with nuclear counterstain (blue), Imaris section mode. The centrally directed processes of the Type 1 niche cells outline the lumen of the uncompressed cell-free cavity (arrows). The star marks the nucleus of an enlarged first generation neural progenitor (NP). Note the extended unlabeled cytoplasm of the NP extending into the stream to the right. (e) Semi-schematic depiction of one half of the adult DPS. One of the Type 1 cells is highlighted in yellow to illustrate its bipolar extensions toward the niche cavity and the neuron clusters. Further abbreviations: ol, olfactory lobe [Color figure can be viewed at wileyonlinelibrary.com] In the studies presented here, we examined more than 500 DPS preparations of Procambarus clarkii in order to shed more light on the cellular basis and migration patterns of the adult-born NP lineages.
Samples were labeled for structural cytoskeletal proteins (tubulin and/or F-actin) as well as different combinations of markers for cell proliferation or mitosis. Special focus was put on the niche and migratory streams in search for consistently recurring cell patterns. We minimized artifacts caused by physical brain sectioning or tissue distortion via mounting of complete brains without z-axis compression, thus maintaining the natural spatial arrangements in the DPS. This was followed by documentation and detailed analysis of virtual threedimensional (3d) stacks of more than 200 of the 500 DPS preparations. Using this setup, we provide evidence for (a) niche cell replen- Facility at room temperature (RT) on a 12/12 light/dark cycle. They were kept in artificial pond water (double-distilled water with added trace minerals and sodium bicarbonate as a buffer) in aquaria and trays equipped with recirculating filter systems. Aquaria and trays contained gravel and some larger stones as well as plastic tubes and a few plastic plants as hiding spots. This enriched environment has been shown to promote high cell proliferation rates in the adult neurogenic system of procambarid crayfish (e.g., Ayub et al., 2011).

| Treatments of experimental animals
The data set analyzed in this study has been compiled from confocal microscopic scans of the DPS obtained during three different years (2015, 2017, and 2018). The animals studied covered varying sizes ranging from 10 mm to 45 mm carapace length (CL; measured from the posterior rim of the orbital cavity to the end of the carapace). Subgroups of these had been exposed to one of the following treatments prior to sacrifice, as they were originally contributing to several different experiments.
Specimens subjected to the last two treatments were part of a series of adoptive transfer experiments in which proliferation marker-

| Dissection and fixation procedures
Irrespective of experiment, all animals were cooled in ice for 5-10 min prior to dissection. Next, the anterior part of the cephalothorax was cut off immediately and placed in cold 4% paraformaldehyde in phosphate-buffered saline (PBS; Boston BioProducts #BM-155). In order to guarantee swift penetration of the fixative into the tissues, the central brain was then removed from the "head" capsule. After 1-1.5 hr fixation at RT with gentle agitation on a horizontal shaker the neural sheath was carefully dissected from the brain, followed by immersion fixation overnight at 4 C. Only dextran-injected brains were not desheathed to avoid damage to delicate vasculature spanning between brain structures and the neural sheath. All brains were processed immunohistochemically after fixation, however, some underwent in-situ hybridization with DIG-labeled riboprobes prior to immunohistochemical labeling. In the present study, only the results of the immunohistochemical labeling procedures are reported.  Nonspecific binding of secondary antibodies was tested by omission of primary antibodies. This resulted in complete loss of specific signal. However, occasionally hemocytes of the granular type (see Lin & Söderhäll, 2011), attaching to the brain's surface or located in the vasculature, exhibited nonspecific cytoplasmic labeling. Additionally, the azide-based labeling reaction for EdU detection occasionally resulted in weak staining of granular cytoplasm in untreated hemocytes. However, these patterns of nonspecific labeling did not impact data interpretation, as most markers of interest were localized in the nucleus (Hoechst, BrdU, EdU, PH3), and also the targeted structural markers with cytoplasmic location (tubulin, F-actin) show specific staining characteristics that are readily distinguished from the granular cytoplasmic hemocyte labeling.
Notably, immunohistochemical processing with the anti-PH3 antiserum (see above) resulted not only in the expected staining of mitotic profiles but also in nonspecific labeling of the cytoplasm of a few single cells (presumably hemocytes; see last results section for more details). It is currently unknown, to which nontarget antigen(s) this antiserum is binding.

| Mounting of samples
After final rinsing in PBS, brains were transferred into nonhardening

| Data documentation, analysis, and presentation
In total, labeling patterns in more than 500 DPS preparations were checked with confocal laser scanning microscopy (cLSM), using a Leica DMI 6000 CS microscope coupled to a Leica TCS SP5 II scan unit.
Out of these, a selection of more than 200 samples with interesting cellular patterns was imaged for 3d analysis. Based on the excitation spectra of the applied fluorochromes, a combination of UV laser (405 nm ! Hoechst), argon laser (488 nm ! Alexa Fluor ® 488) and helium-neon laser (543 nm ! Cy™3; 633 nm ! Alexa Fluor ® 647) was chosen.
The 3d reconstruction software Imaris, V7.0.0 (http://www.bitplane. com/Default.aspx, RRID:SCR_007370) was used for subsequent analyses. Within the Surpass and Section visualization modules of this program (e.g., Figure 1b-d, respectively), software tools were applied as previously described (e.g., Brenneis, Stollewerk, & Scholtz, 2013). Importantly, oblique slicers were used in the Surpass module in order to create a 3d-curved virtual section plane. This enabled visualization of cell arrangements along the curved brain surface in a single 2d projection Supporting Information movies were generated in Imaris (Animation module) and subsequently transformed into MP4-format using the freeware FormatFactory (http://www.pcfreetime.com).

| Cell divisions in the niche and beginning migration of NPs
To visualize the niche and the exact dimensions of its central cavity, nuclear staining was coupled to phalloidin and/or tubulin labeling ( Figure 1c,d). The latter two markers highlight the short processes of the Type 1 cells that form the inner cavity surface. Immunolabeling of the mitosis marker phosphorylated histone H3 was performed only in a subset of specimens (e.g., white arrow in Figure 3e), but advanced stages of M phase could also be reliably identified based on the nuclear staining (white arrows in Figure 3a-d,f-h). Further, tubulin labeling enabled detection of parts of the spindle apparatus or the F I G U R E 2 Creation of 3d-curved virtual section planes in Imaris. Acetylated tubulin and BrdU immunolabeling (white and red, respectively) coupled to nuclear counterstain (blue). Carapace length (CL) of specimen analyzed indicated in (b). (a) Sequence of steps to obtain a virtual section plane that follows the 3d-curvature of migratory stream in an uncompressed whole-mount brain of Procambarus clarkii. Starting from a 3d volume (Imaris Surpass mode) of the scanned portion of the migratory stream, oblique slicers are successively oriented along its course on the 3dcurved ventral surface of the accessory lobe. To control for correct slicer position, the 3d volume view can be deactivated while the slicers are being positioned. This allows flexible visualization of the structures included in the curved section that is being created and enables fine adjustment of slicer orientation. (b) 2d projection of the created 3d-curved virtual section (compare Figure 7e). A clone of four BrdU-positive NPs (one in telophase) is clearly visible as it migrates in the stream toward neuron Cluster 10 [Color figure can be viewed at wileyonlinelibrary.com] midbody that still spans between two sister cells prior to their final abscission (e.g., solid white arrowheads in Figure 3f-h,a,c-h).
While each of the more than 500 DPSs (in~250 brains) processed in the course of the study has been checked with cLSM, ultimately not all of them were imaged and documented as virtual 3d stacks. In total, 3d stacks of 219 niches were obtained ( Table 2). One of the imaging criteria for niches was the presence of mitotic profiles. Accordingly, the 47 niches with mitoses included in the niche data set (Table 2) represent approximately 10% of the DPSs checked in total.
F I G U R E 3 Legend on next page.
3.1.1 | Morphologically symmetric divisions occur next to the central niche cavity Using the Imaris section mode, we analyzed the location and orientation of mitotic profiles in the niche data set in more detail (Figure 3a).
In animals of all sizes, we found mitotic profiles near the central cavity (Table 2); these represent about half of the mitoses observed, found in roughly 5% of all DPSs checked. The mitotic profiles are located among the short processes of the Type 1 niche cells, either directly adjacent to the cavity lumen (white arrows in Figure 3a- The division plane of the majority of meta-, ana-, and telophase profiles was found to be oriented perpendicular to the surface of the cavity (81% of all mitoses next to cavity; Table 2), indicating that both sister cells will initially come to lie side by side next to the latter (e.g., Figure

| NPs entering the migratory streams stay aligned with their relatives
About 90% of the more than 500 niches studied did not contain mitotic cells or pre-abscission sister cells with midbody. Nonetheless, NP pairs at a slightly later stage, that is, after initiation of migration into the streams, can be still frequently identified. The nuclei of these NP pairs are spherical to oval and typically larger than the Type 1 cell nuclei (NP nuclei labeled with stars in Figure 5a-e). As might be expected for actively proliferating NPs, they are often positively labeled in specimens that had been exposed to in vivo S phase  Figure 5f).
The NP closer to the cavity often possesses a slightly larger nucleus and a more voluminous cytoplasm (Figures 3j and 5b,c,f) but the observed size differences between both cells varied between samples (compare, e.g., Figure 5a,b,d,e). The NP nucleus further away from the niche was found at varying distances from the latter, suggestive of its active migration into the streams.
In carefully oriented optical sections, it was sometimes possible to trace a cytoplasmic projection of the larger NP all the way to the cavity lumen (Figure 5b). Also along the streams, the extended cytoplasm of the smaller NP could be followed for some distance (Figure 5a,d).
Notably, not every niche contains enlarged nuclei of active NPs (73% of the 3d-documented niches; Table 2). Further, when present, typically only a single enlarged NP or one NP pair aligned along the niche-stream axis is found per niche half. Only rarely, more than one NP/NP pair is encountered in one niche half ( Figure 5f ).

| Migration and expansion of NP lineages along the streams
More advanced stages of NP lineages were found along the migratory streams. The majority of observations were made in the stream leading to Cluster 10 ( Figures 6-8), as it is easily observable and follows a comparably straight course on the ventral side of the accessory lobe.
However, similar processes seem to occur in the stream toward Cluster 9 that curves around the root of the antenna 1 nerve and is not as readily visualized (Figure 9).

| Multicellular clones of NPs migrate along the streams
In a specimen exposed to EdU (5 days survival time), we identified a labeled NP in anaphase at the niche-stream border, together in a single widened track with another EdU-positive cell further distally in the stream (Figure 6a). This shows that after the first NP division in the niche, the daughter cell closer to the cavity may divide again before migration into the streams. Such additional divisions would be expected to generate linearly arrayed cell clones in the streams, comprising three cells or even more (in case of multiple divisions). Indeed, such strings of multiple cells can be observed (Figures 6b-f and 7), frequently showing proliferation marker labeling in specimens that were exposed to BrdU or Edu over several days (Figures 6d,e and 7c-e; Movie S1). Distances between the NP nuclei within linear clones were found to vary. In some cases, the two NPs nearest to the niche are close to each other, often with signs of nuclear asymmetry, which points to their recent division (Figure 6b,c,f; Movie S1). In other cases, their two nuclei are well-separated and of oval shape, suggestive of faster migration of the more distal cell after a division (Figures 6d,e and 7b). In the distal stream (closer to the proliferation zone associated with neuron Cluster 10), however, cells of a clone may be more tightly packed (Figure 7a-d). These varying patterns indicate T A B L E 2 Descriptive statistics of different niche characteristics analyzed. The category "advanced mitotic profiles" subsumes meta-, ana-, and telophases

| Multiple clones can migrate at the same time along the streams
We also encountered more complex cell patterns in some of the streams studied (Figure 8). Their 3d analysis revealed the simultaneous occurrence of more than one linear cell clone along a migratory stream.
In these cases, different clones occupy separate "lanes" in the stream, positioned either side by side (Figure 8a Table 2).
Also among the additional niches that were checked with cLSM but eventually not imaged (>500 DPS processed in total), no nuclei were found in the cavity lumen. Given this considerable number of niches analyzed, this finding is unlikely to be a sampling artifact and suggests that hemocytes do not enter into the cavity lumen under basic conditions, regardless of animal size or the treatment used in the current studies.

| Adoptive transfer experiments result in proliferation marker-labeled cells in the niche
To further study and confirm direct interactions between hemocytes and the DPS under basic conditions, we performed a series of adoptive transfer experiments, in which hemolymph of BrdU-or EdU- whereas the remaining two niches were found in one specimen sacrificed 3 weeks after transfer. Notably, these niches did not contain single but multiple weakly labeled nuclei in their outer cell layers as well as in their core (Figure 10a,b). The morphology of these labeled cells is not distinguishable from adjacent unlabeled Type 1 niche cells, featuring the typical nuclear size and shape. intensely Hoechst-stained (Figure 11a,b). Occasionally, the nucleus may also have a conspicuous blunt end on one side (e.g., Figures 9d   and 11d). Based on these nuclear characteristics, such cells could be identified even without the nonspecific cytoplasmic labeling in the streams (yellow arrowheads in Figures 6c,f, 7e, 8a, 9a,c,d and 11d), where they occur mostly singly-rarely also in pairs-but never in conspicuous linear arrays as the NP clones. Further, these cells also did not label for proliferation markers in any of our experiments (e.g., Figures 7e and 11d), indicating that they are either cycling very slowly or already postmitotic. To date, the exact nature of this cell type is unclear.
Interestingly, however, cells with identical nonspecific cytoplasmic labeling and nuclear morphology were also found near the surface of the DPS (Figures 3e and 11a-c) and more widely dispersed in other brain areas (not shown). In dextran-injected animals, some of them were revealed to be located inside the arterioles around the niche (red arrowheads in Figure 11b,c), which points toward hemocytes circling in the brain's vasculature. Additional tentative support for this notion may come from the study of the hemolymph of P. clarkii. Among the different hemocytes, small elongated cells with short processes, an intensely labeled nucleus and a relatively homogeneous nongranular cytoplasm are found (Figure 11e).

| Intrinsic cell divisions contribute to the replenishment of the niche cell pool
Previous studies have shown that only few in vivo proliferation marker-labeled cells are found in the niche of P. clarkii at any given time point within the first post-exposure days (e.g., Benton et al., 2011;Song et al., 2009;Sullivan et al., 2007). Further, mitotic profiles have been observed even more rarely (Song et al., 2009;Sullivan et al., 2007;Zhang et al., 2009) and when captured in a late stage of mitosis (ana-or telophase), they were located close to the niche-stream boundary with a division plane perpendicular to the niche-stream axis (Zhang et al., 2009). This is in line with the NP divisions described here. However, we found additional mitoses next to the cavity lumen among the short processes of Type 1 cells that have not been documented before. Their location opposite to the stream and the orientation of their division planes perpendicular to the surface of the cavity lumen speaks against an immediate migration of the two prospective daughter cells into the streams. Moreover, the occurrence of these divisions in animals of all size groups studied (with a possible trend of a size-dependent decrease, Table 2) and across different experimental setups prior to sacrifice (Table 1) demonstrates that their presence is neither restricted to a certain animal age nor strictly treatment-dependent. Instead, they seem to be a characteristic feature of the DPS in adult crayfish. The observation that these mitotic cells extend processes toward the cavity and into the stream indicates that they represent cycling bipolar Type 1 niche cells.
As the niche is composed of more cells with increasing animal size (Zhang et al., 2009), there needs to be a mechanism driving this rise in niche cell numbers. Apart from the attraction and integration of extrinsic cells ; see below), intrinsic multiplication via niche cell divisions is another possible mechanism that has, however, not been previously observed (Zhang et al., 2009). The mitoses next to the cavity described here fit such an intrinsic cell amplification mechanism, being locally separated from the differently oriented neurogenic divisions of the first generation NPs. Accordingly, we propose that the symmetric Type 1 cell divisions adjacent to the cavity constitute one of the processes by which the niche cell pool is replenished in the crayfish DPS (see below, Figure 13).

| The crayfish DPS: A pseudostratified neuroepithelium with INM-like dynamics?
Similarities between these replenishment divisions next to the cavity and cell divisions in the neuroectoderm of other metazoan taxa become more apparent, when recognizing the epithelial-like nature of the crayfish DPS. On first glance, this epithelial-like character is masked by the extreme elongation of the Type 1 cell processes and the clustering of the majority of nuclei in the niche region ( Figure 12a).
However, the cavity is surrounded by the short Type 1 cell processes, which in an ultrastructural investigation have been shown to be tightly

| Bipolar niche cells differentiate into proliferating first generation NPs
In addition to mitoses near the cavity that contribute to the replenishment of the niche cell pool, we also confirmed divisions of the enlarged first generation NPs with cleavage planes oriented roughly perpendicular to the niche-stream axis (e.g., Song et al., 2009;Zhang et al., 2009). However, deviating from previous reports (Zhang et al., 2009), we found that late mitotic phases and cytokinesis of these NP divisions are not restricted to the niche-stream boundary but may also occur centrally in the niche cell cluster4. We could further confirm that first generation NPs actively in the cell cycle can be morphologically identified based on their larger, more spherical nucleus and a more voluminous cytoplasmic compartment than the surrounding Type 1 niche cells (Song et al., 2009). The less intense labeling of the large NP nuclei with fluorescent nuclear stains suggests predominantly euchromatic as opposed to heterochromatic DNA. Interestingly, all of these features fit well with the histological and ultrastructural features of Type 2 niche cells (Zhang et al., 2009;Chaves da Silva et al., 2012), which we propose to represent the active first generation NPs in the niche.
In contrast to the findings of Song et al. (2009) and to results in spiny lobsters and non-reptant decapod crustaceans (Schmidt, 2007;Schmidt & Derby, 2011;Wittfoth & Harzsch, 2018), we did not detect an enlarged NP in every crayfish DPS studied. In fact, a substantial portion of our 3d-analyzed samples lacked this morphologically conspicuous cell type (Table 2). Together with previous results of proliferation marker retention studies Sullivan et al., 2007), this contradicts theories of a permanently enlarged neural stem cell (i.e., a large adult neuroblast) at the base of adult-born lineages in crayfish. In our current study, the cell projections extending from dividing first generation NPs and their two daughter cells toward the cavity and into the streams favor instead their derivation from bipolar Type 1 cells that have entered the cell cycle. This is also consistent with the separate "lanes" along which the different cell clones migrate through the DPS streams: with a Type 1 cell that differentiates into an active first generation NP as

| Migrating clones encompass different cell generations and include transit-amplifying intermediate NPs
String-like arrangements of NPs as they migrate toward the proliferation zones have been previously documented (e.g., Sullivan et al., 2007;Zhang et al., 2009), but the common lineage of cells linearly aligned along the streams was not recognized. Prior to the current study, the model for the adult NP lineages in crayfish assumed that three NP generations occur spatially separated along the DPS (e.g., Benton et al., 2011Benton et al., , 2014Zhang et al., 2009). Our new data lead to several modifications of this model.
Notably, this morphological asymmetry can be very pronounced in some taxa, whereas it is less obvious in others (e.g., Harzsch, 2001;Hein & Scholtz, 2018) and may additionally decrease with ongoing development (e.g., Homen & Knoblich, 2012). Further, while morphological asymmetry of two newly forming NP daughter cells is one indicator for their non-equivalence, it does not per se suffice to prove long-term self-renewal of the parent NP, as expected for a bona fide stem cell. In the case of the DPS in adult crayfish, we concur with earlier findings that no long-term self-renewing stem cells reside in the niche . This view is supported by (a) the absence of an enlarged cell (= potential adult neuroblast) in a significant portion of the niches studied, (b) the observed cell division and migration patterns, and (c) the absence of in vivo cell proliferation marker retention in the niche cell pool for even few days after a marker pulse (e.g., Benton et al., 2011Benton et al., , 2014. Instead, the entire progeny of a first generation NP exits the niche. But in contrast to the previous DPS model, the first generation NPs may divide more than once prior to their niche exit (see Figure 6a) and very likely continue to do so for some time as they migrate at the tail end of expanding clones (proximal to the niche). Accordingly, the tailend NPs rather qualify as a type of intermediate NP (INP) with limited short-term proliferative potential. This is reminiscent of the specialized NP lineages that form the central complex in the brain of Drosophila and grasshoppers, where transit-amplifying INPs with short-term self-renewal capacity lead to a significant amplification of these lineages compared to other neural lineages in the brain (Bello, Izergina, Caussinus, & Reichert, 2008;Boyan & Williams, 2011;Homen & Knoblich, 2012 To date, however, we are unable to determine, if all daughter cells of a tail-end INP divide again and whether they do so according to a stereotyped mode (e.g., just once like typical embryonic ganglion mother cells; see Ungerer & Scholtz, 2008). Nevertheless, the observation of simultaneously migrating clones next to each other in the streams and their deviating cell numbers speaks strongly for a spatiotemporally flexible rather than an invariant cell division pattern. Moreover, not only the cell number but also the distance between nuclei and their shape can show considerable variations between and even within clones (see also Zhang et al., 2009). This suggests temporally dynamic nucleus/cell migration speeds, with similarities to saltatorial nucleokinesis during neuronal migration in vertebrates (Marín, Valiente, Ge, & Tsai, 2010;Tsai & Gleeson, 2005). As a consequence of these various dynamic processes, cells observed at any given time point in the streams are likely to represent a mixture of different NP generations that can belong to separate adult NP lineages.
3. In the proliferation zones within the neuron clusters, we are still unable to reliably discriminate between different adult NP lineages with the available morphological toolkit. This basal-most region of the crayfish DPS frequently features mitotic profiles, which have been previously assigned to third generation NPs (Zhang et al., 2009). In our revised DPS model, it might be more fittingly said to be a clone accumulation area, in which different NP generations of still expanding lineages become spatially packed prior to cell cycle exit and neuronal differentiation.

| Where do hemocytes integrate into the DPS?
Interactions between the DPS and hemocytes of the innate immune system have been a focus in previous studies. Injection of fluorescently labeled dextran into the dorsal sinus revealed an intimate F I G U R E 1 3 Different mechanisms of niche cell pool replenishment. Continuous cell loss in the niche due to the migration of activated first generation neural progenitors (NPs) and their expanding lineages (highlighted in vermillion) can be compensated for by different mechanisms. See Figure 12 for orientation of the neuroepithelium. (a) Intrinsic replenishment in the niche cell pool. Type 1 niche cells (highlighted in yellow) undergo symmetric replenishment divisions. The M phase of these divisions occurs next to the cavity lumen (= apical side of the hypothetical planar neuroepithelium), which is reminiscent of interkinetic nuclear migration (INM) in pseudostratified epithelia. Both daughter cells remain in the niche cell cluster, but may later differentiate into NPs that give rise to a migrating NP lineage. (b) Niche cell pool replenishment via an extrinsic source (modified from Benton et al., 2011Benton et al., , 2014. Hemocytes (most likely of the semi-granular type) attach and integrate into the niche cell cluster, where they differentiate into Type 1 cells (highlighted by blue shading) and may later on turn into enlarged NPs. (c) Niche cell pool replenishment via extrinsic source and intrinsic divisions. Attachment and integration of hemocytes into the niche cell clusters (blue shading) coupled to symmetric replenishment divisions (yellow) counteract and even overcompensate for the depletion of the niche cell pool to provide cellular material for life-long NP specification and migration (vermillion). Potentially, hemocyte-derived Type 1 cells undergo one or more symmetric replenishment divisions as well (to date, the exact sequence of divisions is still unknown). Further, single hemocytes may show different types of interactions with the DPS and progress as single cells through the streams [Color figure can be viewed at wileyonlinelibrary.com] spatial association of the niche and brain vasculature and even indicated direct communication of the hemolymph with the niche cavity (Sullivan et al., 2007;Chaves da Silva et al., 2012. Moreover, in-vitro experiments showed that one hemocyte type (semi-granular cells) is selectively attracted to and attaches to the niche . This intriguing finding was then further explored in vivo via adoptive transfer experiments. Here, hemolymph of proliferation marker-exposed donor animals was injected into untreated recipient animals, which resulted in the appearance of labeled cells in recipient niches, migratory streams and Clusters 9 and 10 after different posttransfer time periods, coupled to their neurotransmitter expression in the clusters which is indicative of neuronal differentiation . Up to now, however, it is unclear which local area of the niche acts as gateway for hemocyte integration and the cavity has been discussed as one of the possible options  increases in cells associated with the niche have been observed when the system is perturbed, for example by exposure to increased levels of serotonin or astakine . The mechanisms by which hemocytes are attracted to the niche and integrate into the cell cluster under varying conditions are not known and await the ability to examine these questions with live-imaging approaches.
In spite of the low success rate of our adoptive transfer experiments, the niches that did show positive proliferation marker labeling clearly support such hemocyte integration into the niche (see also Benton et al., 2014). Notably, we found multiple labeled cells with typical nuclear morphology of Type 1 cells in the niche cell   (Brenneis et al., 2013;Brenneis & Scholtz, 2014;Brenneis, Scholtz, & Beltz, 2018;Döffinger, Hartenstein, & Stollewerk, 2010;Dohle, 1964;Pioro & Stollewerk, 2006;Stollewerk, 2004). These internalized epithelial cell clusters continue to produce additional cells for the differenti-ating central nervous system. Notably, the invagination processes leading to the clusters affect neither the spatial arrangement of neighboring cells to each other nor the apico-basal polarization laid down during embryonic ectoderm development. As a consequence, the previously outward facing apical cell poles are directed toward the center of the clusters and converge around a tiny central cavity in an arrangement very similar to the crayfish niche. In some sea spiders, fibrous bundles of elongate cell processes also emanate from these neurogenic clusters into the soma cortex and have been shown to act as a migration pathway for newborn cells into the ganglion (Brenneis et al., 2018;Brenneis & Scholtz, 2014). Given these structural and functional similarities to the crayfish DPS, it would not be surprising if the latter was likewise formed via ectodermal invagination in late embryonic stages, with a retention of the apico-basal polarization. So far, however, the developmental origin of the DPS remains elusive, as the available studies started either too late in development (Song et al., 2009) or were not focused on the potential relationship between apical embryonic ectoderm and emerging DPS (Sintoni et al., 2012).
4.8 | Cell fate switch as in the Drosophila optic lobe and the mammalian neocortex?
In the simplified planar DPS model, the spindles of the apical replenishment divisions are aligned in parallel to the epithelial surface (Figure 13a,c).
This closely resembles the symmetric proliferative divisions of neuroepithelial cells in the optic lobe of Drosophila and the developing mammalian cerebral cortex (e.g., Brand & Livesey, 2011). Likewise, the proposed transformation of Type 1 niche cells into the active first generation NPs is similar to these two model systems, in which progeny of symmetrically dividing epithelial cells transform into repeatedly dividing NPs (neuroblasts and radial glial cells, respectively). As in the crayfish DPS, this is accompanied by a change of their spindle orientation (from parallel to oblique/perpendicular to the surface). In mammals, this changed spindle orientation has been even implicated to cause the switch from proliferative divisions to the neurogenic NP division mode (Xie, Jüschke, Esk, Hirotsune, & Knoblich, 2013). However, in the optic lobe of Drosophila, experimentally induced change of spindle orientation is not sufficient for transformation into an NP (Egger, Boone, Stevens, Brand, & Doe, 2007). Instead, transient repression of members of the Notch signaling pathway via the proneural gene lethal of scute of the achaete-scute transcription factor family has been shown to promote NP differentiation in a restricted area of the epithelium (Egger, Gold, & Brand, 2010). This interplay of spatio-temporally tightly regulated proneural gene expression and Notch signaling is an evolutionarily conserved mechanism governing neural commitment of epithelial cells in the development of various arthropod groups (Skeath & Thor, 2003;Wheeler, Carrico, Wilson, Brown, & Skeath, 2003;Kux, Kiparaki, & Delidakis, 2013 [insects]; Stollewerk et al., 2001;Stollewerk, 2002 [spiders]; Dove & Stollewerk, 2003;Kadner & Stollewerk, 2004 [myriapods]) and many other taxa, including vertebrates (see Hartenstein & Stollewerk, 2015). Accordingly, a similar molecular switch may be key for the selection of single Type 1 niche cells and their differentiation into first generation NPs in the crayfish DPS. Gene expression studies that seek to explore this aspect are currently underway.

| Parallels with neurogenesis in the adult mammalian brain
In the adult mammalian brain, neurogenesis occurs in two principal regions, the subventricular zone (SVZ) in the wall of the lateral ventricle and the subgranular zone (SGZ) of the dentate gyrus. Adult neurogenesis was initially believed to be sustained by long-term self-renewing NPs that reside in the both regions (B1 cells in the SVZ; radial glia-like cells [R-cells] in the SGZ) and undergo predominantly asymmetric divisions in stem cell-like manner. In the meantime, however, this has turned into a matter of debate (e.g., Bond, Ming, & Song, 2015) and an increasing body of evidence indicates considerable heterogeneity among the primary adult NPs, many of which appear to lack a capacity for long-term selfrenewal and become quickly exhausted once activated (Calzolari et al., 2015;Encinas et al., 2011;Fuentealba et al., 2015;Pilz et al., 2018). A recent study now even suggests that neurogenesis in the SVZ is not even driven by asymmetric divisions of the B1 cells (Obernier et al., 2018). Instead, the majority of them were observed to undergo symmetric consuming divisions that generate two transit-amplifying NPs and thus deplete the B1 cell pool, whereas the remaining smaller fraction passes through symmetric self-renewal, producing two B1 cells. Similarly, a live-imaging study on the SGZ has documented only limited selfrenewal of R-cells via asymmetric but also symmetric divisions, prior to their exhaustion in consuming neurogenic divisions (Pilz et al., 2018).
Notably, asymmetric self-renewal divisions were observed to follow on symmetric ones but never vice versa, thus showing similarities to cortical development during which NPs irreversibly transition from a symmetric proliferative to an asymmetric neurogenic phase.
These novel insights into the cellular dynamics of mammalian adult neurogenesis feature again some remarkable parallels to our proposed crayfish DPS model (Figure 13c), where bipolar Type 1 niche cells may undergo symmetric proliferative divisions before transforming into first generation NPs that in turn enter consuming neurogenic divisions and give rise to clones of migrating NPs with finite proliferation potential.
These parallels between the life-long neurogenic mechanisms of two taxa as phylogenetically distant as crayfish and mammals may serve as another illustration of how evolution independently generates similar solutions when faced with similar challenges. Seen from this perspective, the recently discovered novel mode of NP pool replenishment via integration and neural differentiation of hemocytes in the crayfish DPS  may well represent a more ubiquitous pathway than originally appreciated and thus inform the future direction of studies on mammalian systems (see Beltz et al., 2015;Beltz, Brenneis, & Benton, 2016 for further discussion). grant-nos. BR5039/1-1, BR5039/3-1). The project was supported by a National Science Foundation grant to BSB (NSF-IOS-1656103).

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.