Neuronal sources of hedgehog modulate neurogenesis in the adult planarian brain

The asexual freshwater planarian is a constitutive adult, whose central nervous system (CNS) is in a state of constant homeostatic neurogenesis. However, very little is known about the extrinsic signals that act on planarian stem cells to modulate rates of neurogenesis. We have identified two planarian homeobox transcription factors, Smed-nkx2.1 and Smed-arx, which are required for the maintenance of cholinergic, GABAergic, and octopaminergic neurons in the planarian CNS. These very same neurons also produce the planarian hedgehog ligand (Smed-hh), which appears to communicate with brain-adjacent stem cells to promote normal levels of neurogenesis. Planarian stem cells nearby the brain express core hh signal transduction genes, and consistent hh signaling levels are required to maintain normal production of neural progenitor cells and new mature cholinergic neurons, revealing an important mitogenic role for the planarian hh signaling molecule in the adult CNS. DOI: http://dx.doi.org/10.7554/eLife.19735.001


Introduction
Once thought to be a non-existent phenomenon, homeostatic adult neurogenesis is a common trait shared by many disparate organisms including rodents, birds, flies, and humans (Altman, 1962(Altman, , 1969Doetsch et al., 1999;Goldman and Nottebohm, 1983;Eriksson et al., 1998;Fernández-Hernández et al., 2013). However, levels of neuronal turnover are tightly limited in these animals (Obernier et al., 2014). In fact, the highest known site of adult homeostatic neurogenesis in the human central nervous system (CNS) is the hippocampus, where annual neuronal turnover rates are estimated to be only 1.75% (Spalding et al., 2013;Sanai et al., 2011;Bergmann et al., 2012). Adult neurogenesis in most animals depends on the action of ectodermally derived neural stem cells, which have radial-glial character and are integrated into a stable niche microenvironment. Extrinsic signals such as wingless (Wnt), sonic hedgehog (Shh), and bone morphogenetic proteins (BMPs) act to finely control neural stem cell proliferation and differentiation (Silva-Vargas et al., 2013;Lehtinen et al., 2011).
The asexual strain of the freshwater planarian, Schmidtea mediterranea (S. mediterranea), is a constitutive adult flatworm (Lophotrochozoan) that challenges the dogma of low rates of adult neurogenesis. Not only does S. mediterranea constantly turnover its brain, but also it is capable of complete brain regeneration within only two weeks following decapitation (Cebria, 2007;Reddien and Sánchez Alvarado, 2004;Newmark and Sánchez Alvarado, 2002). In addition, the uninjured planarian CNS is known to be a highly dynamic organ, which can adjust its size through the addition or subtraction of mature neurons to maintain consistent proportions with the rest of the body as it grows and shrinks, respectively (Baguñá and Romero, 1981 ;Hill and Petersen, 2015). Amazingly, these regenerative feats and high levels of homeostatic neurogenesis are accomplished in the absence of a recognizable neuroepithelium, and without any definitive neural stem cells (van Wolfswinkel et al., 2014;Zhu et al., 2015). Recently, brain-derived Wnt signals have been shown to influence the neurogenic output of planarian stem cells (neoblasts) during regeneration (Hill and Petersen, 2015). However, little is known about the specific extracellular signals and transcription factors that modulate neoblast activity within this body region to balance cell proliferation and neuronal differentiation, which undoubtedly involves many overlapping regulatory systems.
Here, we have identified two planarian homeodomain transcription factors, Smed-nkx2.1 and Smed-arx (henceforth referred to as nkx2.1 and arx), which act to specify cholinergic, GABAergic, and octopaminergic neurons within the ventralmedial (VM) region of the planarian CNS. Interestingly, these very same VM neural cell types are a major source of the planarian hedgehog ligand (Smed-hh; henceforth referred to as hh) (Rink et al., 2009;Yazawa et al., 2009). We also describe a novel neoblast population, which are nkx2.1 + or arx + , express the core machinery for the reception of hh and Wnt signals, and are located adjacent to hh + VM neurons. Finally, we present evidence that consistent hh signaling within this neoblast microenvironment is required to promote normal homeostatic neurogenesis of the VM neuronal population. In total, we identify a hh signaling axis that positively modulates VM neurogenesis through distinct progenitor cells.
eLife digest Most animals can continue to generate and add new neurons in their nervous system into adulthood, though the process is often tightly regulated. In adult humans, only a small number of neurons are made or lost, such that the fewer than 2% of the neurons in the nervous will change over, or "turnover", the course of a year. The turnover of neurons in some other animals is much higher than it is in humans. A freshwater flatworm, called Schmidtea mediterranea, is one example of such an animal that can even regenerate an entirely new brain if its head is decapitated. These flatworms have a large population of adult stem cells, which makes these high rates of neuron production and regeneration possible. However, it is largely unknown if this population contains stem cells that can only become new neurons, in other words "dedicated neuronal stem cells". Moreover, it is also not clear what kinds of signals communicate with these stem cells to promote the production of new neurons.
In animals from flies to humans, a signaling molecule encoded by a gene called hedgehog forms part of a signaling pathway that can promote neuron production during development. Therefore, Currie et al. asked if the hedgehog signaling molecule might communicate with the stem cells in adult flatworms to control how many new neurons they produce.
The experiments revealed that the hedgehog signaling molecule is almost exclusively produced by the flatworm's brain and the pair of nerve cords that run the length of the flatworm. Currie et al. then found a smaller group of cells close to the flatworm's brain that looked like dedicated neural stem cells. These cells can receive the hedgehog signals, and further experiments showed that flatworm's brain requires hedgehog signaling to be able to produce new neurons at its normal level.
The hedgehog signaling molecule is likely only one of many signaling molecules that regulate the production of new neurons in flatworms. It will be important to uncover these additional signals and understand how they work in concert. In the future, a better understanding of this process will help efforts to devise ways to induce humans to replace neurons that are lost following injury or neurodegenerative diseases. DOI: 10.7554/eLife.19735.002 Video 1. Control(RNAi) worms exhibiting normal locomotion and ability to flip over from dorsal surface back onto ventral surface. DOI: 10.7554/eLife.19735.003 Results nkx2.1 and arx are expressed in ventral-medial neural cell types nkx2.1 and arx were originally cloned and isolated during an RNAi screen aimed at identifying planarian transcription factors with potential roles in neuronal specification. A unique behavioral defect was observed in all nkx2.1(RNAi) animals, characterized by tonic muscular contractions that bend the head dorsally, so that it is perpendicular with the rest of the body. Interestingly, some of these nkx2.1(RNAi) worms end up on their lateral body edge, causing these animals to move in tight circles, compared to control(RNAi) animals which move in straight lines on their ventral surface (Videos 1 and 2). In contrast, arx(RNAi) animals do not display overt behavioral defects, except when flipped onto their dorsal side where these worms have difficulty to corkscrew in order to reorient onto their ventral surface compared to control(RNAi) and wild-type animals (Videos 1 and 3).
In order to investigate the root cause of these behavioral defects, and whether nkx2.1 and arx might function in neuronal specification and maintenance, whole-mount in situ hybridizations (WISH) and cross-sections were first performed on wild-type animals. nkx2.1 exhibited broad expression throughout the body but was clearly present within the CNS (black arrow), whereas arx expression was spatially restricted to the medial CNS, including the anterior brain and the full length of the ventral nerve cords ( Figure 1A). Transverse sections through the anterior-posterior (A-P) midpoint of the brain enabled imaging in greater spatial detail and highlighted the concentrated expression of arx within the ventral-medial region of each brain lobe ( Figure 1B). In addition, expression of nkx2.1 and arx was observed within a few isolated cells in between the brain lobes, a region largely populated by adult stem cells ( Figure 1B, yellow arrows).
In order to determine which neuronal subtypes express nkx2.1 and arx, double fluorescent in situ hybridization (dFISH) was performed, comparing either of these two transcription factors with established markers of either cholinergic, GABAergic, octopaminergic, dopaminergic, or serotonergic neurons (Nishimura et al., 2007a(Nishimura et al., , 2007b(Nishimura et al., , 2008a(Nishimura et al., , 2008b(Nishimura et al., , 2010. Both nkx2.1 and arx were found to be expressed by neural cell types that occupy the VM region of the planarian brain (cholinergic, GABAergic, and octopaminergic neurons) but were largely excluded from dopaminergic and serotonergic neurons ( Cholinergic neurons are by far the most populous cell type in the planarian CNS and are densely distributed throughout the brain ( Figure 1C), making it difficult to assess the level of expression of nkx2.1 and arx within this entire neuronal population. Therefore, a more focused region of interest was imaged to quantify the degree of expression of nkx2.1 and arx within cholinergic neurons (chat + ) (Nishimura et al., 2010). Transverse sections were first taken at the A-P midpoint of the brain, and individual brain lobes were imaged. Each lobe was then divided into a VM and dorsal-lateral (DL) half, with the line of division occurring between the widest point of each brain lobe ( Figure 1C; yellow line). A considerable proportion of cholinergic neurons expressed nkx2.1 in both the VM (36.89 ± 5.46%) and DL (23.33 ± 5.13%) brain regions ( Figure 1C). In contrast, arx expression was more heavily biased toward the VM brain region, where 63.21 ± 4.35% of cholinergic neurons expressed this transcription factor ( Figure 1C).
Due to the relatively modest size of the octopaminergic neuronal subpopulation (tbh + ), and the focal grouping of these cells within the medial region of the CNS, this neural subtype was assessed as a whole. Interestingly, expression of both transcription factors was observed within this neural cell group, with 76.9 ± 6.66% of octopaminergic neurons exhibiting nkx2.1 expression, while a lower proportion were arx + (33.44 ± 6.54%) ( Figure 1E). Taken together, these expression data suggested a potential role for nkx2.1 and arx in the specification and/or maintenance of VM cholinergic, GABAergic, and octopaminergic neurons.

nkx2.1 and arx are required for the maintenance of VM neural cell types
In order to test the functional roles in neuronal specification, nkx2.1 and arx were knocked down by RNAi, and the cholinergic, GABAergic, and octopaminergic neuronal subtypes were assayed for any changes to cell number during homeostasis only (see Materials and methods). Compared to control (RNAi) animals, a loss of arx resulted in a visible reduction of cholinergic neurons in the VM region of the CNS (Figure 2A; white arrows), while nkx2.1(RNAi) worms appeared relatively unaffected (Figure 2-figure supplement 1). In order to quantify the loss of cholinergic neurons in arx(RNAi) animals, transverse sections of single brain lobes were imaged as described above, and 20 mm of confocal depth was recorded. Subsequently, the number of chat + cholinergic neurons were counted within each of the VM and DL brain regions. Strikingly, arx(RNAi) worms exhibited a near 40% reduction in the number of VM cholinergic neurons, whereas DL neurons were largely unchanged ( . This very specific loss of VM cholinergic neurons in arx(RNAi) worms, combined with the evidence that arx was expressed by >60% of these neurons, demonstrated that arx was required for the maintenance of VM cholinergic neurons.
nkx2.1 represented an excellent candidate to act in the specification of VM GABAergic neurons based on its expression within nearly all of these cells. As there are many fewer GABAergic than cholinergic neurons in the planarian brain, imaging and quantification of this entire neuronal population was performed. As expected, arx(RNAi) animals exhibited no loss of GABAergic neurons in either the VM or the DL subgroups ( Figure 2B Interestingly, both nkx2.1(RNAi) and arx(RNAi) animals exhibited a significant reduction in the number of octopaminergic neurons, which correlated with the degree of expression for each transcription factor within this neural cell type ( Figure 2C,F). While nkx2.1(RNAi) animals exhibited a > 60% reduction in octopaminergic neurons, arx(RNAi) worms displayed a more modest but significant~25% reduction ( Figure 2C,F). It should also be noted that the neurons lost in nkx2.1 (RNAi) or arx(RNAi) animals were located in the VM region of the planarian brain ( Figure 2C). As expected, RNAi of nkx2.1 or arx had no appreciable effect on the dopaminergic and serotonergic neuron populations (Figure 2-figure supplement 2). Therefore, the maintenance roles of nkx2.1 and arx are limited to the VM cholinergic, GABAergic, and octopaminergic neuronal cell populations.

hedgehog is expressed in VM neural cell types
Previous studies have shown that the planarian CNS is a major site of expression for the single hedgehog ligand in planarians (Smed-hh; henceforth referred to as hh) ( Figure 3A) (Rink et al., 2009;Yazawa et al., 2009). However, it is unknown what specific cells express this signal and what cells respond to it. Transverse sections through the brain highlighted hh expression within the VM regions of each brain lobe, similar to that of arx ( Figure 3B). More specifically, hh expression was observed in the same VM neural cell types as described above for nkx2.1 and arx. As with arx, hh expression within cholinergic neurons was heavily biased toward the VM region of the brain lobe, where 23.17 ± 2.88% of these neurons expressed the signaling molecule ( Figure 3C). In addition, hh was expressed by nearly all VM GABAergic neurons (98.08 ± 3.85), as well as a considerable proportion of octopaminergic neurons (33.7 ± 2.23%) ( Figure 3D

arx(RNAi) leads to a reduction of hh expression from VM neurons
Although only partially overlapping, the finding that hh was co-expressed within the same VM neural cell types as nkx2.1 and arx ( Figure 3C-E) suggested that these two transcription factors may have upstream regulatory function on neuronal hh expression. WISH experiments demonstrated a clear reduction in hh expression within the medial CNS of arx(RNAi) worms ( Figure 3F; white arrows), while silencing nkx2.1 had little effect on hh expression (Figure 3-figure supplement 2). This reduction in hh expression was quantified in the same manner described above for cholinergic neurons, by imaging transverse sections through single brain lobes, and counting the number of hh + cells in the VM and DL brain regions, within 20 mm of confocal depth. This analysis revealed a near 60% reduction in the number of VM hh + neurons in arx(RNAi) animals ( Figure 3F-G), but no significant loss of hh + cells in the DL brain ( Figure 3-figure supplement 2). In contrast, silencing of nkx2.1 did not significantly alter hh expression levels in either brain region ( Figure 3G, Figure 3figure supplement 2). While it is unclear whether arx directly regulates hh expression in VM neurons, or whether reduced hh expression is related to a loss of VM cholinergic neurons, it is clear that hh expression in the VM domain of the planarian brain requires arx gene function.
Single-cell RNA sequencing confirms neuronal expression of hh, nkx2.1, and arx Recent studies have begun using single-cell RNA sequencing (scRNAseq) technologies, to investigate the complete transcriptional profiles of individual planarian stem cells and mature cell types (Wurtzel et al., 2015;Molinaro and Pearson, 2016). From the scRNAseq dataset generated by Wurtzel et al., all cells from uninjured animals (including stem cells, neurons, gut, epithelial, muscle, and parapharyngeal cells) as well as neurons from injured worms were analyzed for expression of hh. From this group of sequenced cells, hh expression was detected in 4/39 neurons, and a single stem cell, but no other cell types ( Figure 3H), supporting previous data that hh expression is largely restricted to the nervous system.
In addition, these 39 sequenced neurons were examined for expression of nkx2.1 and arx, which were detected in two and six neurons, respectively, with two of these cells displaying co-expression with hh ( Figure 3I). Expression of nkx2.1 or arx within hh + neurons in the brain was also directly observed by dFISH (Figure 3-figure supplement 3). High expression levels of common mature neuron markers (Smed-synaptotagmin, Smed-synapsin, Smed-pc2) were found in all nine sequenced cells, and more specifically, five of these cells were chat + cholinergic neurons ( Figure 3I) (Agata et al., 1998;Cebrià, 2008). In addition, each of these nine neurons expressed a unique combination of neuropeptides (Collins et al., 2010), transcription factors (Cowles et al., 2013(Cowles et al., , 2014Currie and Pearson, 2013;Pineda et al., 2002), RNA-binding proteins (Higuchi et al., 2008;Koushika et al., 1996), and even 2 Wnts with known or putative roles in neuronal specification and patterning ( Figure 3I). This in silico data acts to further confirm that the planarian hh ligand is expressed by mature neurons.

Planarian neoblasts expresses hh signal transduction machinery
Considering that VM planarian neurons are a major source of the hh signaling molecule (Figure 3) (Rink et al., 2009;Yazawa et al., 2009), it was of great interest to discover what cells and tissues might be the target of this brain-derived signaling molecule. Previous observations have shown that hh signaling activity can influence global proliferation levels (Rink et al., 2009). Therefore, the adult stem cells (ASCs) that occupy the space in between the two brain lobes were examined as a potential target of planarian hh.
Reception and effective transduction of the hh signaling pathway requires the receptors Patched (ptc) and Smoothened (smo) as well as the Gli transcription factors (Ho and Scott, 2002). Confirming previous reports, WISH experiments for the planarian homologs of these key signal transduction effectors (Smed-ptc, Smed-smo, and Smed-gli-1) (Rink et al., 2009) demonstrated relatively ubiquitous expression throughout the body of the adult animal, with Smed-gli-1 showing a particularly strong signal within the gastrovascular system ( Figure 4A). However, when piwi-1 + ASCs (Reddien et al., 2005b) were visualized in between the two brain lobes, many were observed to coexpress the ptc receptor ( Figure 4B). Similarly, smo and gli-1 expression was detected in brain-adjacent ASCs ( Figure 4B), providing support for the notion that planarian stem cells might be a target of brain-derived hh signals.
In order to support these dFISH findings, we examined gene expression from a recent dataset (Molinaro and Pearson, 2016) containing 96 single stem cells in the head (isolated from a FACS gate termed X1) as well as 72 cells from a FACS gate termed X2 that contains a mixture of stem cells and their immediate division progeny (Zhu et al., 2015;Hayashi et al., 2006). These cells were examined for expression of ptc, smo, or gli-1. Individual neoblasts were first examined for expression of established stem cell markers and cell cycle regulators (piwi-1, piwi-2, piwi-3, Smed-bruli, Smed-tdrd1L2, Smed-chd4, Smed-mcm2, Smed-pcna, Smed-cyclinB-2) (Zhu et al., 2015;van Wolfswinkel et al., 2014;Guo et al., 2006;Reddien and Sánchez Alvarado, 2004). These stem cells were then binned as either X1 or X2 and arranged in descending order based on ptc then smo then gli-1 expression ( Figure 4C). Approximately 30% (56/168 cells) expressed at least one of ptc, smo, or gli-1 ( Figure 4C). Interestingly, approximately half of these sequenced cells (81/168) also expressed at least one of the nine planarian Frizzled (fz) receptors (which bind Wnt ligands) (Gurley et al., 2008(Gurley et al., , 2010. These in silico findings were also confirmed by performing dFISH experiments, which showed expression of all nine fz receptors within piwi-1 + ASCs in between the two brain lobes (  (Wurtzel et al., 2015). Taken together, these dFISH and in silico data suggested that planarian stem cells adjacent to the CNS are one of the cellular targets of brain-derived hh and Wnt signals.
A subset of nkx2.1 + and arx + neoblasts and post-mitotic progenitors exists in close proximity to the VM brain region Recent studies have highlighted an interesting property of planarian transcription factors: they are expressed not only by the cells and tissues that they act to specify, but also within ASCs nearby the mature organ in question (Currie and Pearson, 2013;Cowles et al., 2013;Scimone et al., 2011Scimone et al., , 2014Reddien, 2011, 2012). Therefore, ASCs nearby the VM brain region were investigated for expression of the nkx2.1 and arx transcription factors. By examining transverse sections through the brain, stem cells, which express piwi-1 mRNA and PIWI-1 protein, can be observed in great numbers, and are largely excluded from the two brain lobes ( Figure 5A-F). In addition, their post-mitotic progenitors, which shut down piwi-1 expression, but briefly maintain detectable amounts of PIWI-1 protein, can be visualized in this region (Guo et al., 2006;Wenemoser and Reddien, 2010). Some of these progenitors can be found several cell diameters within the mature brain lobes ( Figure 5A-F; green arrowheads). In this specific spatial region, expression of nkx2.1 or arx was observed within a significant number of stem cells or post-mitotic progenitors ( Figure 5A-F). Quantification of piwi-1 + stem cells revealed that a significant percentage displayed expression of nkx2.1 (7.99 ± 0.95%), whereas arx expression was less frequent (1.74 ± 0.99%). Quantification of all PIWI-1 + cells (encompassing both stem cells and their post-mitotic progeny) demonstrated an expansion of nkx2.1 and arx expression, to include 13.29 ± 1.96% and 8.11 ± 1.83%, respectively, of all cells counted.

Hedgehog signaling regulates neural progenitor dynamics
Based on the expression of hh within VM neurons and the hh-transduction machinery within adult stem cells, the hh signaling pathway was next assessed for a potential role in the regulation of neural progenitor cells and the ultimate production of new neurons. This was performed by RNAi knockdown of the hh ligand, resulting in decreased signaling activity, or by knockdown of the ptc receptor (a negative regulator of pathway activity), resulting in increased hh signaling levels (Rink et al., 2009). Neural progenitor cells were then visualized by taking transverse sections through the brain, and imaging PIWI-1 + stem and progenitor cells in between the brain lobes for expression of several transcription factors of putative neural stem/progenitors (nkx2.1, arx, lhx1/5-1, coe, and pax6a) ( Figure 6A) (Currie and Pearson, 2013;Cowles et al., 2014;Pineda et al., 2002;Scimone et al., 2014). hh(RNAi) and ptc(RNAi) animals were investigated for changes to the percentage of PIWI-1 + stem/progenitor cells expressing these neural transcription factors. Silencing of the planarian hh ligand resulted in a significant reduction of expression for all five neural transcription factors within PIWI-1 + cells ( Figure 6B). In contrast, levels of neural progenitor cells were largely unchanged in ptc (RNAi) animals, with the exception of coe and pax6a which exhibited a small but significant increase in PIWI-1 + cells ( Figure 6B).
The thymidine analog bromodeoxyuridine (BrdU), which is incorporated into dividing stem cells during DNA synthesis and subsequently chases into their post-mitotic progeny, has proven key for planarian lineage-tracing experiments (Newmark and Sánchez Alvarado, 2000), some of which  have revealed that the intact planarian CNS maintains high levels of continuous neurogenesis (van Wolfswinkel et al., 2014;Zhu et al., 2015). Therefore, in order to investigate whether hh signaling levels regulate neuronal homeostasis and the production of new cholinergic neurons, a BrdU pulse-time chase experiment was performed. Following gene silencing of hh or ptc, RNAi-treated animals were exposed to a single pulse of BrdU. After 6-and 9-day chase periods, newly generated cholinergic neurons were detected by FISH for chat expression as well as nuclear BrdU staining ( Figure 6C-E). The level of neurogenesis was then quantified by counting the number of chat + BrdU + cells observed within 30-mm-thick confocal stacks from within the planarian brain. hh(RNAi) animals consistently displayed significantly lower cholinergic neural homeostasis both after 6-day (~35% reduction) and 9-day chase periods (>40% reduction), while silencing of ptc had little effect ( Figure 6C-F). To investigate whether this decrease in neuronal homeostasis levels affected the intact nervous system, long-term RNAi experiments were performed (10 feedings), and all five neuronal populations were examined. Of particular interest were the VM cholinergic, GABAergic, and octopaminergic neural populations, which express the hh ligand ( Figure 3C-E). However, decreased hh signaling did not affect the intact population of any neuronal cell type ( Figure 6-figure supplements 1 and 2; see Discussion).
To determine whether brain-derived hh signals were required specifically to maintain normal levels of neurogenesis, or whether this ligand acts globally to stimulate the production of all planarian cell types, progenitor cells and BrdU incorporation were examined for four additional tissues; the eyes, pharynx, epidermis, and gut. Importantly, these other planarian tissues were largely unaffected by perturbations to hh signaling, except for the eyes, which exhibited reduced progenitor cell numbers in hh(RNAi) animals, and increased BrdU incorporation after ptc knockdown ( Figure 6-figure  supplements 3 and 4). These combined experiments, demonstrated that neuronal sources of the planarian hh ligand specifically regulate their own homeostasis through distinct progenitors in the uninjured brain.

Discussion
Planarians not only constantly turnover their CNS at relatively high rates, but must also be able to respond to massive injury to regenerate an entire brain de novo. This work has demonstrated that two planarian homeodomain transcription factor homologs, nkx2.1 and arx, function in the longterm maintenance of cholinergic, GABAergic, and octopaminergic neurons in the VM region of the planarian CNS ( Figure 2). Surprisingly, these same three VM neural cell types were also found to express the hh signaling molecule ( Figure 3C-E). Neuronal hh expression was also confirmed by using a scRNAseq dataset that included mature neurons (Wurtzel et al., 2015), and additionally revealed that all four of the sequenced hh + neurons co-expressed a planarian Wnt ligand, Smed-wnt11-6 ( Figure 3I). Specialized nkx2.1 + or arx + neoblasts were also identified, which were typically localized adjacent to the VM region of the planarian CNS, and express key members of the hh and Wnt signal transduction machinery ( Figure 5). Finally, the planarian hh pathway was found to be required to maintain steady levels of neurogenesis in the intact brain, as a reduction in hh signaling activity, led to decreased production of neural stem/progenitor cells and newly generated cholinergic neurons ( Figure 6G).
Similar to planarian muscle cells, many of which express important signaling molecules that maintain positional control along the body axis and during regeneration events (Witchley et al., 2013;Scimone et al., 2016), mature planarian neurons express hh and Wnt ligands which regulate neurogenesis levels and maintain patterning (Hill and Petersen, 2015). Previous studies have shown that the Wnt ligand Smed-wnt11-6 and Wnt antagonist Smed-notum are expressed by cholinergic neurons, and that this Wnt signal acts on planarian stem cells to repress neuron production and prevent posterior expansion of brain tissue during regeneration (Hill and Petersen, 2015). Interestingly, we show that several planarian neurons simultaneously express this repressive Wnt ligand as well as the hh signaling molecule, which promotes the production of neurons ( Figures 3I and  6G). Conceptually, it makes sense that signals originating from the mature CNS, would act to repress neurogenesis programs. This would help to establish a steady-state system, which modulates neurogenesis levels depending on the size of the existing CNS. However, our observations suggest that contrary to this model, neuronal sources of hh actively promote neurogenesis levels in both the brain and of eye photoreceptors neurons ( Figure 6-figure supplements 3 and 4).
It is unclear, exactly how the planarian hh signaling pathway achieves rates of neuronal homeostasis, whether it acts as a permissive cue, influencing neoblasts adjacent to the brain to produce neural cell types, or as a mitogenic signal, to control proliferation levels of neural-biased stem cells. Similar to mechanisms of Hh action in other systems, the regulatory loops are likely to be complex in planarians due to the fact that long-term hh(RNAi), surprisingly, did not produce robust neural deficits over a span of five weeks ( Figure 6-figure supplements 1 and 2). Perhaps with body-wide changes in proliferation in hh(RNAi), rates of neural cell death are also decreased, leading to little change in neuronal populations (Rink et al., 2009). Similarly, in the mammalian CNS, expression of Shh from differentiating and mature neurons is required to maintain normal proliferation levels of embryonic neural precursors and adult neural stem cells, respectively (Á lvarez-Buylla and Ihrie, 2014;Ihrie et al., 2011;Fuccillo et al., 2006).
Unlike most organisms where neural stem cells are packed within an organized neuroepithelium, planarian neural-biased stem cells are situated in the mesenchymal space in between the two brain lobes ( Figure 5). While this loose grouping of stem cells offers few similarities to a true neuroepithelium in terms of structure or cellular origin, the local signaling microenvironment may fulfill a similar function. Computationally, neural-committed stem cells, termed nNeoblasts, were recently detected and may be the targets of brain-derived signals (Molinaro and Pearson, 2016). However, while nkx2.1 + and arx + stem cells exhibit a spatial distribution and transcriptional profile suggestive of neuronal lineage commitment ( Figure 5G), without definitive lineage-tracing studies, these neoblasts cannot yet be classified as true neural stem cells.
In chordate nerve cord development, Hh is expressed ventrally, along the long body axis and has strong ventralizing roles in cell fate determination (Briscoe and Ericson, 2001;Jessell, 2000). This is not typical of arthropods (Matise, 2007;Arendt and Nübler-Jung, 1999). It is interesting to speculate that VM neural fates in planarians are specified using mechanisms more associated with chordates. In support of this idea, the role of nkx2.1 and arx in the maintenance of planarian cholinergic and GABAergic neurons was particularly noteworthy (Figure 2), as their mammalian counterparts (Nkx2.1 and Arx) are known to fulfill similar roles and also be under the control of sonic hedgehog (Jessell, 2000;Colasante et al., 2008). In the embryonic rodent ventral telencephalon, Nkx2.1 acts in upstream neural progenitor cells to produce both cortical GABAergic interneurons and striatal cholinergic interneurons (Butt et al., 2008;Lopes et al., 2012;Sussel et al., 1999), whereas Arx functions downstream in the terminal differentiation and migration of cortical GABAergic interneurons (Vogt et al., 2014). Notwithstanding this slight deviation in cell fate determinism, it appears that these two transcription factors have retained a remarkable degree of functional conservation across this significant evolutionary gap.

Materials and methods
Cloning Smed-nkx2.1 and Smed-arx were found by homology to mammalian orthologs in the sequenced and assembled planarian genome and transcriptomes as previously described (Currie and Pearson, 2013;Robb et al., 2008). Primers were designed and full length genes were cloned by 3' RACE. Smed-nkx2.1 is transcript number dd_Smed_v6_13898_0_1 from (http://planmine.mpi-cbg.de/, Brandl et al., 2016) and Smed-arx was deposited to Genbank under the accession number KX961610.

Animal husbandry and RNAi
Asexual S. mediterranea CIW4 strain were reared as previously described (Sánchez Alvarado et al., 2002). RNAi experiments were performed using previously described expression constructs and HT115 bacteria (Newmark et al., 2003). Briefly, bacteria were grown to an O.D.600 of 0.8 and induced with 1 mM IPTG for 2 hr. Bacteria were pelleted and mixed with liver paste at a ratio of 500 ml of liver per 100 ml of original culture volume. Bacterial pellets were thoroughly mixed into the liver paste and frozen as aliquots. The negative control, 'control(RNAi)', was the unc22 sequence from C. elegans as previously described (Reddien et al., 2005a). All RNAi food was fed to 7-day starved experimental worms every third day for six total feedings, and fixed 10-14 after the final RNAi feeding. All animals used for immunostaining were 2-3 mm in length and size-matched between experimental and control worms.
Immunolabeling, in situ hybridizations (ISH), and BrdU WISH, dFISH, and immunostaining were performed as previously described (Currie et al., 2016). Colorimetric WISH stains were imaged on a Leica M165 fluorescent dissecting microscope. dFISH and fluorescent mouse-anti-PIWI-1 (gift from Dr. Jochen Rink) stains were imaged on a Leica DMIRE2 inverted fluorescence microscope with a Hamamatsu Back-Thinned EM-CCD camera and spinning disc confocal scan head. Cell counts and co-localizations were quantified using freely available ImageJ software (http://rsb.info.nih.gov/ij/). Significance was determined by a two-tailed Student's ttest. All experiments were, at minimum, triplicated and at least five worms were used per stain and per time point. All labeling images were post-processed using Adobe Photoshop.

Analysis of scRNAseq data
Raw scRNAseq data from head-specific X1 and X2 cells from (Molinaro and Pearson, 2016) can be downloaded under NCBI Gene Expression Omnibus (GEO accession GSE79866). Raw scRNAseq data from uninjured cells (including stem cells, neurons, gut, epithelial, muscle and parapharyngeal cells) as well as neurons following injury generated by Wurtzel et al., were obtained from the NCBI Sequence Read Archive (SRA:PRJNA276084). Reads were aligned to the SmedASXL transcriptome assembly under NCBI BioProject PRJNA215411 using bowtie2 (Langmead and Salzberg, 2012) with 15 bp 3' trimming. The read count data were log2-transformed (log2(count + 1)), violin plot was produced using modified source code from (Macosko et al., 2015) and the heatmaps were produced using the modified heatmap.3 source code from (Molinaro and Pearson, 2016).