Transcription factors lhx1/5-1 and pitx are required for the maintenance and regeneration of serotonergic neurons in planarians

The form and function of any individual neuron is determined by its very specific gene-expression signature: the combinatorial expression of particular enzymes, receptors and structural proteins defines its unique neural identity (Hobert, 2011; Hobert et al., 2010). This neural identity is coordinated and maintained throughout the lifespan of the neuron by the expression of terminal selector transcription factors, and in some cases a single terminal selector gene can regulate the expression of hundreds of downstream effector genes (Eastman et al., 1999; Etchberger et al., 2007; Flames and Hobert, 2009; Hobert, 2011; Liu et al., 2010; Uchida et al., 2003). The instructive power of these regulatory factors has become increasingly clear with recent studies showing that the ectopic expression of terminal selector genes can directly reprogram germ cells and fibroblasts into fully differentiated neurons (Patel et al., 2012; Tursun et al., 2011; Vierbuchen et al., 2010). Considering the relatively poor regenerative abilities of the human central nervous system (CNS), there is a great therapeutic interest in the ability to convert a patient’s own skin cells into missing neural cell types using specific combinations of transcription factors. However, before such practices are possible, we require greater understanding of how terminal selectors function during neural regeneration and how these regenerating neurons integrate themselves into the existing neuronal circuitry. To this end, highly regenerative model systems excel to allow dissection of the molecular mechanisms underpinning regeneration and cell-fate specification.

The freshwater planarian Schmidtea mediterranea (S. med), is a free-living flatworm of the lophotrochozoan superphylum and is well known for its ability to regenerate any missing body part as an adult, including the entire CNS and peripheral nervous system (PNS) (Cebrià, 2007; Cebrià, 2008; Cebrià et al., 2002; Gentile et al., 2011; Reddien and Sánchez Alvarado, 2004). This regenerative ability is made possible by a large population of pluripotent adult stem cells (ASCs) (Sánchez Alvarado and Kang, 2005; Wagner et al., 2011). S. med is a constitutive asexual adult and is amenable to gene-function studies by use of RNA interference (RNAi), which is fed to adult animals and subsequently spreads to all tissues, including neurons (Sánchez Alvarado and Newmark, 1999). Therefore, planarians are an ideal model system to study the molecular mechanisms of neuronal function, specification and regeneration, all in an adult animal in vivo. Multiple studies have previously shown that the planarian nervous system contains common invertebrate neuronal subtypes, which exist in specific ratios in the brain and PNS (Agata et al., 1998; Cebrià et al., 2002; Nishimura et al., 2007; Nishimura et al., 2008b; Nishimura et al., 2008c). However, it remains unknown what factors specify each neuronal subtype, how patterning mechanisms during regeneration generate precise subtype ratios, or what the function of each neuronal subtype may be in the larger context of the animal’s behavior.

Here we have described the function of two homeodomain transcription factors, Smed-lhx1/5-1 and Smed-pitx (henceforth referred to as lhx1/5-1 and pitx), which were required to regulate the maintenance and regeneration of planarian serotonergic neurons. We found that lhx1/5-1 and pitx were expressed in all fully differentiated serotonergic neurons, where they act to maintain the expression of multiple canonical serotonergic marker genes that designate the unique identity and function of this neural cell type. We also show that specifically during regeneration, lhx1/5-1 and pitx were co-expressed with the stem cell marker smedwi-1 in a novel progenitor cell. When either lhx1/5-1 or pitx was knocked-down by RNAi, we observed the rapid loss of functional serotonergic neurons. Surprisingly, we observed that the loss of serotonergic neurons blocked the ability of ventral motile cilia to drive planarian locomotion. We further showed that the locomotion defect was characterized not by the loss of motile cilia, but instead by the loss of metachronal synchrony across the ventral epidermis. Together, these results showed that serotonergic neuronal inputs can be used across large distances to coordinate and tune the beating of cilia in order to produce directed movement. Finally, we demonstrated that in addition to its role in serotonergic neurons, pitx was also required during regeneration to specify the planarian midline by controlling Smed-slit expression in the anterior and Smed-wnt1 in the posterior.

Lhx and Pitx gene family homologs were found by homology in the sequenced and assembled planarian genome, as previously described (Pearson and Sánchez Alvarado, 2010; Robb et al., 2008; Sánchez Alvarado et al., 2002; Sánchez Alvarado et al., 2003). Primers were designed and full-length genes were cloned by 3′ rapid amplification of cDNA ends (RACE). Cloned open reading frames (ORFs) were then converted to predicted proteins and subjected to Bayesian phylogenetic analyses for Lhx family members. Protein sequences used in phylogenies were obtained from the NCBI Entrez protein database or directly from the genome sequencing projects of included organisms. The program Geneious (www.geneious.com) was used with the MUSCLE alignment plugin and two tree building plugins for Geneious were used as independent analyses. Both maximum likelihood and Bayesian analyses were performed with the following settings: maximum likelihood - 10,000 bootstrap replicates, WAG substitution model, estimated distances; Bayesian - 1.1 million replicates, WAG substitution model, 4 heated chains, 25% burn-in, subsample frequency of 1000. Consensus tree images were saved through Geneious, which were then manipulated in Adobe Photoshop. Alignment of the Pitx homeodomain is shown in supplementary material Fig. S3. Smed-lhx1/5-1, Smed-pitx, Smed-tph, Smed-aadc, Smed-vmat, Smed-sert, Smed-kchan, Smed-hmctn, Smed-upps, Smed-asic and Smed-sh3bp2 sequences were deposited in GenBank as accession numbers: KF134112-KF134122. Transcripts listed in supplementary material Tables S1-S3 can be found in the transcriptome database associated with the NCBI GEO project: GSE47348.

Asexual individuals of S. med 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 OD₆₀₀ of 0.8 and induced with 1 mM IPTG for 2 hours. Bacteria were pelleted and mixed with liver paste at a ratio of 500 μl liver per 100 ml 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 five total feedings. Nomenclature denotes time of staining and number of feeds (e.g. 5fd10 is equal to 10 days after the fifth feed). Amputations were performed 3 days after the final feeding unless noted otherwise. All animals used for immunostaining were 2-3 mm in length and size-matched between experimental and control worms.

RNA deep sequencing (RNA-seq) was performed on whole animals 12 days following RNAi feedings (5fd12) for: control(RNAi), lhx1/5-1(RNAi) and pitx(RNAi). Experiments were performed in experimental triplicate and sequenced to a depth of ∼50 million reads per sample, and multiplexed on an Illumina HiSeq2000 with 50 bp single-end reads. Sequences were then aligned to a previously described planarian transcriptome with the program Bowtie2 (Labbé et al., 2012; Langmead et al., 2009). To determine significantly up- or downregulated transcripts between experimental animals and controls, the freely available baySeq software was used (http://www.bioconductor.org/packages/2.11/bioc/html/baySeq.html) (Hardcastle and Kelly, 2010). Full sequences and results were submitted to the NCBI bioproject PRJNA205281.

Whole-mount in situ hybridization (WISH), double-fluorescence in situ hybridization (dFISH) and immunostaining were performed as previously described (Lauter et al., 2011; Pearson et al., 2009). Colorimetric WISH stains were imaged on a Leica M165 fluorescent dissecting microscope. dFISH and fluorescent anti-α-tubulin (Iowa Hybridoma Bank; 12G10) 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 colocalizations were quantified using freely available ImageJ software (NIH). Significance was determined by a two-tailed Student’s t-test. 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.

To visualize beating cilia, live worms were first placed on a microscope slide with a small amount of planarian culture water, and a coverslip with Vaseline on each corner was placed directly onto the worm to immobilize the specimen during video recording. A Leica DMRA2 compound microscope was used to visualize beating cilia along the lateral body edge of the planarian head region under DIC optics at 60× magnification. Videos were captured at 180 frames per second using a Flare 4M180CCL digital camera (IO Industries, www.ioindustries.com) and were edited with Virtualdub (www.virtualdub.org) to adjust for brightness/contrast, crop the specific region of interest and decompile videos for the isolation of individual jpeg images. In order to determine ciliary beat frequency, a small rectangular region of interest perpendicular to the lateral body wall was selected from each video, and this was used to generate kymographs using ImageJ software (NIH) and methods as previously described (Dentler et al., 2009). From the kymograph images, the number of cycles that individual motile cilia are able to complete during the selected time frame of the video segment were counted, and the beat frequency was calculated for that particular specimen.

In order to disrupt serotonin signaling, the serotonin antagonist mianserin hydrochloride (Sigma) was added to wild-type worms at a final concentration of 20 μM. Mianserin-treated worms began displaying defective cilia-based locomotion within seconds after exposure (supplementary material Movie 7). Additionally, the addition of ectopic serotonin hydrochloride (Sigma) at a final concentration of 100 μM was able to interrupt normal cilia-based locomotion within minutes after exposure. This same concentration of serotonin hydrochloride was used in the RNAi rescue experiments to intermittently restore cilia-driven locomotion in lhx1/5-1(RNAi) and pitx(RNAi) worms.

Planarian serotonergic neurons have been described as spatially restricted to the ventral side of the worm with their axonal projections extending across the mediolateral body axis (Fraguas et al., 2012; Nishimura et al., 2007). In order to confirm that this neuronal population is truly serotonergic and conserved at a molecular level with other species, the planarian homologs of four canonical genes required for the biosynthesis, packaging and reuptake of the serotonin neurotransmitter were cloned. Smed-tryptophan hydroxylase (tph) and Smed-amino acid decarboxylase (aadc) are the enzymes that catalyze the conversion of the amino acid tryptophan to the intermediate 5-hydroxytryptophan (5-HTP), and ultimately into serotonin (5-HT) (Deneris and Wyler, 2012; Nishimura et al., 2007). Smed-vesicular monoamine transporter (vmat) packages these serotonin molecules into presynaptic vesicles that are ready for release into the synaptic cleft, and the Smed-serotonin transporter (sert) recycles these released serotonin molecules back into the presynaptic terminal to be repackaged (Fig. 1C) (Deneris and Wyler, 2012).

tph, aadc, vmat and sert displayed similar gene expression patterns to each other by WISH, characterized by punctate staining throughout the body of the worm. It should be noted that large secretory peripharyngeal cells also expressed tph, aadc and vmat in the middle of the animal (Fig. 1A). However, the ventral restriction of these markers was easily visible when viewed in transverse sections taken through the medial region of the planarian brain (Fig. 1A). dFISH demonstrated the overlapping expression of tph with aadc, vmat and sert, confirming that these four genes were all expressed in mature, tph⁺, serotonergic neurons (Fig. 1B). Whereas tph, aadc and sert specifically marked serotonergic neurons, vmat was also expressed by dopaminergic and octopaminergic neurons, as expected (supplementary material Fig. S1) (Nishimura et al., 2008a; Nishimura et al., 2007).

The Lhx and Pitx homeodomain transcription factors have well-established roles in vertebrate neuronal differentiation, and therefore represented good candidates for having a similar function in the planarian nervous system (Miquelajáuregui et al., 2010; Pillai et al., 2007; Smidt et al., 2004; Zhao et al., 2007). Predicted homologs to these gene families were first identified in the S. med genome based on homology to the homeodomain sequence for Pitx and the homeodomain plus LIM domains that are characteristic of Lhx proteins. Eleven planarian lhx genes were identified and, based on phylogenetic analysis, representatives of all six major subfamilies were found to exist in planarians (supplementary material Fig. S2). Only one planarian homolog of the pitx gene family was identified based on the characteristic lysine residue at position 50 of the homeodomain, which is a hallmark of the pitx and bicoid-like transcription factors (supplementary material Fig. S3) (Eastman et al., 1999; Lamonerie et al., 1996).

WISH analyses showed that lhx1/5-1 and pitx both exhibited a punctate expression pattern in ventrally located cells, similar to the distribution of serotonergic neurons (Fig. 1D). dFISH analyses for the serotonergic marker tph and either lhx1/5-1 or pitx confirmed their expression in this mature neural cell type (245/246 tph⁺ neurons also expressed lhx1/5-1 (99.59%); 193/194 tph⁺ neurons also expressed pitx (99.48%) (Fig. 1E,F). As expected, dFISH for both lhx1/5-1 and pitx showed that the two transcription factors also had overlapping expression patterns with each other, in which 79.9% of all pitx⁺ cells were lhx1/5-1⁺ and 39.5% of all lhx1/5-1⁺ cells were pitx⁺ when quantified in the posterior region of the animal (Fig. 1G; supplementary material Fig. S9). It should be noted that a significant amount of lhx1/5-1 and pitx expression was outside of the serotonergic cell compartment. It is unknown what these additional, nonserotonergic cells represent, but at least some are cholinergic and GABAergic neurons (supplementary material Fig. S4). Nonetheless, these results showed that virtually all serotonergic neurons expressed both lhx1/5-1 and pitx transcription factors, making these ideal candidates to regulate aspects of serotonergic identity or function.

To understand perturbations in the serotonergic neuronal population, it was first important to establish the amount of normal cell turnover (homeostasis) for this cell type. When planarians are irradiated with a lethal dose of γ-irradiation (60 Gray), all stem cells are eliminated within 48 hours and no more differentiated cell types can be produced (Bardeen and Baetjer, 1904; Eisenhoffer et al., 2008; Reddien et al., 2005b). Consequently, irradiated worms are unable to survive much longer than 2 weeks. Following irradiation, we observed no detectable changes in the serotonergic neuron population over the course of 2 weeks (supplementary material Fig. S5). This result showed that planarian serotonergic neurons are a stable, long-lived cell type, and that any changes to this cell population within 2 weeks would not be the result of physiological cell turnover.

To determine whether RNAi to lhx1/5-1 or pitx would disrupt the serotonergic neuron population, uninjured worms were fed RNAi five times and then were examined 10 days later (5fd10) (Fig. 2A) for any changes in their expression of the four described markers of serotonergic neurons (Fig. 1A). Following RNAi, the expression of tph and aadc in the ventrally located serotonergic neurons was completely lost in lhx1/5-1(RNAi) and pitx(RNAi) worms, whereas the changes in expression of vmat and sert were less dramatic but still significantly reduced (Fig. 2B). In order to quantify the reduced expression of vmat and sert in RNAi-treated worms, dFISH experiments were performed to accurately count the number of cells expressing these two markers in relation to tph. In the post-pharyngeal tail region of control(RNAi) worms, an average of 54.8±4.6 vmat⁺ cells were observed, with the majority being vmat⁺tph⁺ serotonergic neurons (35/54.8) and the remaining ∼20 cells being vmat⁺tph^-. However, in the post-pharyngeal tail region of lhx1/5-1(RNAi) or pitx(RNAi) worms we observed a significant drop in the average number of total vmat⁺ cells to 26.5±2.1 and 22±5.7 cells respectively, all of which were vmat⁺tph^- (supplementary material Fig. S6). In the tail region of control(RNAi) worms we observed on average 48 sert⁺ cells, all of which were tph⁺. By contrast, in lhx1/5-1(RNAi) worms we only observed ∼12 sert⁺ cells on average, all of which were tph^-, and in pitx(RNAi) worms sert expression was completely lost (Fig. 2B; supplementary material Fig. S7). It should be noted that markers for other neural subtypes were not affected in either RNAi condition with the exception of a ventral GABAergic subpopulation in lhx1/5-1(RNAi) worms (supplementary material Fig. S8). These data showed that lhx1/5-1 and pitx transcription factors were required in serotonergic neurons to maintain the identity of this particular neural subtype by coordinating the expression of multiple genes.

Considering that lhx1/5-1 and pitx were responsible for the expression of several serotonergic neural marker genes, we next tested whether these transcription factors were also required for the maintenance of generic neuronal identity or survival. We took advantage of the fact that lhx1/5-1(RNAi) worms displayed an incomplete knockdown of sert, which allowed us to examine the state of the remaining RNAi-treated serotonergic neurons. In the tail region of control(RNAi) worms, every serotonergic neuron also expressed the pan-neural marker Smed-prohormone convertase 2 (pc2) (46/46), and 41/46 of the sert⁺pc2⁺ neurons displayed visible α-tubulin based axon tracts (Fig. 2C) (Collins et al., 2010; Robb and Sánchez Alvarado, 2002). Similarly in lhx1/5-1(RNAi) worms, all of the sert⁺ RNAi-treated neurons that we observed (28/28) maintained the expression of pc2, and had visible axons (Fig. 2C). These data showed that lhx1/5-1 and pitx function to maintain the specific identity and function of serotonergic neurons, whereas the maintenance of generic neuronal properties are likely to be under the control of other transcription factors.

Because lhx1/5-1 and pitx were both expressed by planarian serotonergic neurons where they functioned to maintain the identity of this neural cell type (Fig. 1G; supplementary material Fig. 2B), we next tested whether a hierarchy exists between these two transcription factors or whether they function independently. In order to test the regulatory relationship between lhx1/5-1 and pitx, lhx1/5-1(RNAi) or pitx(RNAi) worms were examined for any changes to the expression level of the other transcription factor. Although we did not observe any significant change to lhx1/5-1 expression in pitx(RNAi) worms, we observed that the expression level of pitx was dramatically downregulated in lhx1/5-1(RNAi) worms (Fig. 2D). To quantify this result, in the tail region of control(RNAi) worms, we observed an average of 90.6±9.8 pitx⁺ cells, whereas in lhx1/5-1(RNAi) worms this number dropped by 75.93% to 21.8±2.7 pitx⁺ cells on average (supplementary material Fig. S9). By contrast, there was no significant decrease in the number of lhx1/5-1⁺ cells in pitx(RNAi) worms compared to the control(RNAi) condition (supplementary material Fig. S9). It should be noted that complete elimination of pitx expression was not expected because ∼20% of its expression does not overlap with lhx1/5-1 (supplementary material Fig. S9). These data showed that lhx1/5-1 is required for proper pitx expression in serotonergic neurons.

lhx1/5-1 and pitx were required for the long-term maintenance of planarian serotonergic neurons, but it was unclear whether these transcription factors were also involved in the early specification or regeneration of this neuronal subtype. During head regeneration, lhx1/5-1 and pitx expression were observed in head blastemas at 1.5 days after head amputation. By contrast, terminal differentiation markers of mature serotonergic neurons, such as tph, were expressed later and became more pronounced ∼24 hours after lhx1/5-1 and pitx during regeneration (regeneration day 2.5; Fig. 3B). This time difference between the early expression of lhx1/5-1 and pitx and the appearance of terminal differentiation markers suggested that during regeneration, these transcription factors may be turned on in precursor cells that later give rise to mature serotonergic neurons.

In order to test whether immature serotonergic neural precursor cells exist in the head blastemas of wild-type planarians, animals were examined for any lhx1/5-1⁺ or pitx⁺ cells that also expressed known markers of adult stem cells (smedwi-1) (Reddien et al., 2005b) or early stem cell progeny (Smed-prog-1) at 2 days post-amputation (dpa) (Eisenhoffer et al., 2008) (henceforth referred to as piwi-1 and prog-1). Even though serotonergic neurons represent a very small proportion of all planarian cells, we observed measurable piwi-1⁺ cells that also expressed the transcription factors lhx1/5-1 (19/1575 total piwi-1⁺ cells counted) or pitx (10/1294 total piwi-1⁺ cells counted) (Fig. 3C). By contrast, the expression of lhx1/5-1 and pitx was conspicuously absent in prog-1⁺ early progeny cells. Only 2/648 prog-1⁺ cells expressed lhx1/5-1, and 0/517 prog-1⁺ expressed pitx (Fig. 3C). Finally, all newly regenerated serotonergic neurons that expressed the terminal differentiation marker tph, also expressed lhx1/5-1 and pitx as expected (Fig. 3C). Of note, we never observed any piwi-1⁺ stem cells that also expressed the terminal differentiation marker tph (supplementary material Fig. S10), which suggested that the differentiation process from lhx1/5-1⁺pitx⁺piwi-1⁺ precursor cell to lhx1/5-1⁺pitx⁺tph⁺ mature neuron is a distinct process that does not go through a prog-1⁺ progeny stage.

The ventral surface of the planarian epidermis is composed of multiciliated epithelial cells, and in order for normal gliding locomotion to occur, these motile cilia must be formed properly as well as coordinate their beating in ‘metachronal synchrony’ (Glazer et al., 2010; Rink et al., 2009; Rompolas et al., 2010; Salathe, 2007). If adjacent cilia are not coordinated in their beating, they will not produce net forward movement. Previous studies have shown that RNAi to structural components of planarian cilia such as IFT88 or Iguana leads to the loss of cilia and loss of gliding locomotion (Glazer et al., 2010; Rompolas et al., 2010). However, it is largely unknown how large ciliated epithelia in any species are coordinated across large distances, resulting in directional fluid flow, and in the case of planarians, forward gliding.

In addition to the molecular defects in serotonergic maintenance and regeneration, surprisingly, both lhx1/5-1(RNAi) and pitx(RNAi) worms displayed severe locomotion defects (Fig. 4A). In contrast to control(RNAi) worms, which were capable of normal gliding locomotion (supplementary material Movie 1), lhx1/5-1(RNAi) and pitx(RNAi) worms had no gliding ability, yet muscular movements and ‘inch worming’ behavior was unaffected (Fig. 4A; supplementary material Movies 2, 3). We hypothesized that the loss of gliding behavior could either be due to loss of cilia, loss of cilia beating or loss of cilia coordination. Because the ventral and dorsal epithelia meet at the lateral edge of the animal, cilia beating can be imaged from a dorsal view (Rompolas et al., 2010). Using high-speed imaging on control(RNAi) worms, the characteristic wave-like beating pattern of these motile cilia was confirmed, and the ciliary beat frequency (CBF) was calculated to be roughly 25 Hz (Fig. 4B,C; supplementary material Movie 4), which was similar to previously reported CBF measurements in wild-type planarians (Rompolas et al., 2010). When lhx1/5-1(RNAi) or pitx(RNAi) worms were imaged, both displayed normal numbers and length of motile cilia that were also beating. This showed that the observed locomotion defects were not due to defects in ciliogenesis, nor the maintenance of ciliary structures (Fig. 4C). However, following RNAi of either transcription factor, we observed significant disruptions to the synchronous beating patterns of motile cilia as characterized by an individual cilium beating out of phase with a neighboring cilium (Fig. 4C, red arrows; supplementary material Movies 5, 6). In addition to the lack of synchrony observed in each RNAi condition, lhx1/5-1(RNAi) worms did not exhibit any appreciable drop in the CBF (25 Hz), while motile cilia in pitx(RNAi) worms displayed a significantly lower CBF of 15 Hz (Fig. 4B).

Considering that both lhx1/5-1 and pitx were expressed in mature serotonergic neurons, which are spatially adjacent to ventral ciliated epithelial cells, the simplest explanation was that these ciliary locomotion defects could be related to dysfunctional serotonin signaling. In order to examine the influence of serotonin on motile cilia dynamics, we exposed wild-type planarians to mianserin, a pharmacological antagonist that has been shown to block serotonin receptors (Kuang et al., 2002). Within seconds, mianserin-treated worms began to exhibit loss of gliding locomotion defects similar to lhx1/5-1(RNAi) and pitx(RNAi) worms, although muscular inch worming was unaffected (Fig. 4A; supplementary material Movie 7). In addition, mianserin-treated worms displayed uncoordinated motile cilia with a significantly reduced CBF of ∼15 Hz (Fig. 2B,C; supplementary material Movie 8). These data strongly suggested that the ciliary locomotion defects observed in lhx1/5-1(RNAi) and pitx(RNAi) worms were likely to be due to dysfunctional serotonin signaling in neurons rather than defects in cilia or the ciliated epidermis itself. To confirm that ciliary beat frequency coordination depends on serotonin signaling, we hypothesized that too much serotonin would have similar effects on wild-type worms. When serotonin was directly added to wild-type worms, we observed the loss of gliding movement within minutes of addition (supplementary material Movie 9). Surprisingly, when serotonin was added to lhx1/5-1(RNAi) or pitx(RNAi) animals that were incapable of gliding, intermittent rescue of gliding was observed (supplementary material Movies 10, 11). Together, these data demonstrated that serotonergic signaling is essential to coordinate cilia beating to produce gliding behavior in planarians.

In addition to the serotonergic maintenance and locomotion defects in uninjured worms, pitx(RNAi) animals displayed dramatic regeneration defects, resulting in the collapse of the planarian midline and the fusion of normally separate bilateral structures. During normal regeneration of the head and tail, bilateral organ systems such as the anterior eyespots, brain ganglia and the posterior gut branches are kept separated by midline mechanisms (Blassberg et al., 2013; Cebrià et al., 2007; Chen et al., 2013; Gurley et al., 2010; Petersen and Reddien, 2009; Rink et al., 2009). As expected, all control(RNAi) worms regenerated two distinct eyespots (0/33 with cyclopia) in the anterior head blastema, and in the newly formed tail, no cases of gut branch fusion were observed (0/8). By contrast, numerous pitx(RNAi) worms exhibited regenerating eyespots fused at the midline resulting in 14/43 cyclopic worms, and gut branch fusion events were observed in the majority of pitx(RNAi) worms (6/7) (Fig. 5A,B). Finally, pitx(RNAi) worms displayed blunt-end tails with reduced or no apparent posterior blastema (55/56) compared with control(RNAi) worms, in which only one such case was observed (Fig. 5B).

Cyclopia and gut branch fusion phenotypes observed in pitx(RNAi) worms appeared similar to defective slit signaling, which has been shown to function at the planarian midline to keep these tissues separate (Cebrià et al., 2007). In addition, gut branch fusion events have been observed in cases where the expression of posterior wnt genes is disrupted, resulting in abnormal patterns of slit expression (Cebrià et al., 2007; Gurley et al., 2010; Petersen and Reddien, 2009). Therefore, pitx(RNAi) animals were tested for defects in the expression of Smed-slit and the midline tail-specific Smed-wnt1 ligand. During early regeneration (2 dpa), pitx was strongly expressed at the dorsal midline of both the anterior and posterior blastemas, and dFISH experiments showed that the pitx-expressing cells at the anterior pole co-expressed Smed-slit, whereas those at the dorsal pole expressed Smed-wnt1 (Fig. 5C). Consistent with these findings, in pitx(RNAi) worms, the normal midline expression of Smed-slit was significantly decreased in both the anterior and posterior blastemas, and the expression of Smed-wnt1 at the posterior pole was completely lost (Fig. 5D). These results showed an unexpected role for pitx specifically during midline regeneration in planarians, which may represent an ancestral patterning function of the Pitx family of genes, similar to their role in left-right patterning in vertebrates (Campione et al., 1999; Gage et al., 1999).

RNA-seq has recently been applied successfully to planarian studies (Abril et al., 2010; Labbé et al., 2012; Sandmann et al., 2011; Solana et al., 2012). However, it is unknown whether RNA-seq has the sensitivity to detect changes in transcript levels in a cell type that represents so few cells in the entire animal (i.e. several hundred serotonergic neurons in an animal that contains ∼10⁶ total cells). lhx1/5-1 and pitx have both been shown to regulate the identity of planarian serotonergic neurons, but to more broadly understand how they coordinate the expression profile of this neural cell type and to discover additional markers of serotonergic neurons, RNA-seq was performed on lhx1/5-1(RNAi) or pitx(RNAi) worms, and then compared to control(RNAi) worms. From this, 25 transcripts in lhx1/5-1(RNAi) animals and 14 transcripts in pitx(RNAi) animals, were found to be significantly downregulated compared with control(RNAi) worms using baySeq, the most stringent analysis method (Fig. 6A; supplementary material Tables S1-S3). Importantly, lhx1/5-1 and pitx were both 4.5-fold and 14.1-fold downregulated in their own RNAi experiments, respectively. In addition, no known cilia components were significantly changed, which supported the conclusion that cilia were unaffected by lhx1/5-1 or pitx knockdowns. Finally, another positive control for serotonergic neurons, sert, was significantly downregulated in both RNAi conditions (Table 1). The other serotonergic markers used in this study were not significantly changed, which we attributed to their remaining expression in peripharyngeal cells or in the optic cups (Fig. 2B). From both RNAi conditions, we cloned and analyzed a total of 13 genes by WISH in wild type and in RNAi animals. We observed that 2/13 genes were co-expressed in tph⁺ cells (Fig. 6B), and importantly, were both lost after lhx1/5-1 and pitx RNAi (Fig. 6E). These two genes were homologous to hemicentin precursor 1 (hmctn 017595) and a voltage-gated potassium channel (Kchan 017730) (Fig. 6E, Table 1). In addition, we found examples of genes that were only regulated by lhx1/5-1, such as the uroporphyrinogen III synthase (upps 014402) homolog that was expressed specifically in peripharyngeal cells or the acid sensing ion channel 4 (asic4 009950) homolog specifically expressed in ventrally located cells that did not express tph (Fig. 6C, Table 1). An SH3 domain-binding protein 2 (sh3bp2 007007) homolog was expressed in a few isolated cells within the brain region of wild-type worms and was the only gene in our validated clones to be downregulated specifically in pitx(RNAi) animals (Fig. 6D, Table 1).

Due to its remarkable regenerative capabilities, the adult planarian has proven to be an excellent model organism to study the regeneration of specific cell and tissue types from pluripotent somatic stem cells, including the gut, eye and excretory system (Forsthoefel et al., 2011; Lapan and Reddien, 2011; Scimone et al., 2011). However, specific transcription factors that regulate the differentiation of distinct neuronal subtypes in the planarian have not been identified, and thus, it has been difficult to ascribe function to a particular class of neurons. Here we have shown that two homeobox transcription factors, lhx1/5-1 and pitx, are terminal selector genes with important roles in the differentiation and long-term maintenance of planarian serotonergic neurons, which are located throughout the ventral side of the worm adjacent to the ciliated ventral epithelium (Fig. 7A). Similarly to other known terminal selector genes, lhx1/5-1 and pitx function to coordinate the subtype identity and function of serotonergic neurons, but they are not in fact required for their overall pan-neuronal identity (Flames and Hobert, 2009; Hobert et al., 2010; Liu et al., 2010; Uchida et al., 2003). In the absence of either lhx1/5-1 or pitx, serotonergic neurons maintain generic neuronal properties but they lose the expression of important genes required for the biosynthesis and vesicular packaging of the serotonin neurotransmitter (Fig. 7C). Without this crucial input from functional serotonergic neurons, the motile cilia of the ventral epidermis can no longer beat together in metachronal synchrony, and the worms are rendered immobile.

It is known that at least some planarian stem cells are pluripotent, but whether all piwi-1⁺ cells in planarians are true stem cells or restricted progenitor cells has yet to be determined (Guedelhoefer and Sánchez Alvarado, 2012; Wagner et al., 2011). Several studies have recently shown that piwi-1 does indeed overlap with ‘differentiation’ markers for the eye and protonephridia, and that this overlap may represent a progenitor cell that has not fully differentiated but is committed to a particular lineage (Lapan and Reddien, 2011; Scimone et al., 2011). Therefore, it is likely that there may be many unappreciated and dedicated piwi-1⁺ cells that are restricted to a specific cell fate. Our data support this hypothesis because we showed that lhx1/5-1 and pitx overlap with piwi-1⁺ cells specifically when planarians are induced to regenerate serotonergic neurons. Interestingly, we did not observe appreciable co-labeling of lhx1/5-1 or pitx with a known stem cell progeny marker, prog-1 (Eisenhoffer et al., 2008). This suggests that the specification and differentiation of serotonergic neurons in planarians does not proceed through a prog-1⁺ progenitor state (Fig. 7B).

Members of the Pitx gene family have wide-ranging functions in vertebrate developmental processes, including the proper patterning of the hindlimb and pituitary gland, and the differentiation of midbrain dopaminergic neurons (Lamonerie et al., 1996; Smidt et al., 2004; Szeto et al., 1999). Interestingly, one family member, pitx2, is required for both the differentiation of midbrain GABAergic and glutamatergic neurons and also to the nonautonomous establishment of left/right asymmetry in several organ systems (Campione et al., 1999; Waite et al., 2011). Similarly, we have found that the single planarian member of this gene family, Smed-pitx, plays dual roles in the terminal differentiation of neurons as well as the nonautonomous separation of bilateral organ systems during regeneration by regulating the midline patterning genes slit and wnt1 (Fig. 7D). Thus, the multiple functions of Smed-pitx suggests that the ancestral Pitx gene had roles in both neuronal differentiation and organ patterning, which over time have undergone gene duplications and subfunctionalizations during vertebrate evolution.

Many examples of epithelia in the animal kingdom rely on synchronously beating motile cilia to generate fluid flow, but the manner in which these vast fields of tiny beating cilia are instructed to beat together has long remained unanswered (Salathe, 2007). Mathematical modeling has suggested that simple hydrodynamic forces acting between neighboring cilia may be sufficient to self-organize metachronal wave patterns (Elgeti and Gompper, 2013; Gueron et al., 1997; Yang et al., 2008). Although it is likely that such hydrodynamic forces may help promote metachronal synchrony, as this represents an energetically favorable state (Gueron et al., 1997; Yang et al., 2008), our data suggest that the coordination of motile cilia requires additional input in the form of serotonergic neurons in order to generate and maintain metachronal synchrony.

Planarian gliding locomotion can be induced in stationary worms by attraction to food sources and avoidance of light (Hyman, 1951). Both of these behavioral responses are initiated in the anterior cephalic ganglia (brain) and then presumably travel via the ventral nerve cords to reach the proper effector organs. Serotonergic neurons are distributed along the ventral surface of the worm, with their axonal projections largely oriented along the mediolateral axis of the worm, creating a neural network capable of transmitting such a message to the ventral epidermis (Fig. 7A) (Fraguas et al., 2012; Nishimura et al., 2007). The precise circuitry of planarian serotonergic neurons has yet to be determined, but in order to communicate with the outer epithelial tissues it is likely that their axon projections traverse through the multiple muscle layers and a basal lamina, before finally communicating with multiciliated epidermal cells (Fig. 7A) (Hyman, 1951). Although it is unknown how serotonin instructs the synchronous beating of motile cilia, we can hypothesize two scenarios: (1) serotonin might be acting as a permissive cue, where at a specific concentration it induces all motile cilia to beat at the same beat frequency; or (2) the finely tuned and rhythmic release of serotonin maintains a constant frequency of ∼25 Hz (Rompolas et al., 2010). We believe that serotonin concentration is critical due to the fact that both too much and too little serotonin signaling in planarians results in loss of gliding.

While this is the first example of neuronal input being required for the maintenance of metachronal synchrony, serotonin has been previously reported to influence motile cilia dynamics in both invertebrate and mammalian systems. Similar to planarian locomotion, the pond snail Lymnea stagnalis uses a ciliated epidermis to glide across surfaces, and it has been shown that the multiciliated epidermal cells of the snail foot organ are directly innervated with serotonin-containing axon tracts (McKenzie et al., 1998). Direct serotonergic innervation of ciliated epithelial tissues is also observed in the embryo of another pond snail, Helisoma trivolvis, in which serotonin release acts to increase the beat frequency of the associated motile cilia (Koss et al., 2003; Kuang et al., 2002). Similarly, when serotonin is applied to cultured sections of mouse trachea, the motile cilia that line the airway epithelium respond by increasing their ciliary beat frequency, resulting in a faster transport of foreign particles (König et al., 2009). This ciliostimulatory role of serotonin observed in several biological contexts points to a conserved role of this neurotransmitter in the regulation of coordinated ciliary behaviors, and opens the possibility that additional neuronal inputs may be functioning to maintain metachronal synchrony in other ciliated epithelia.

We thank Eric Ross in Dr Alejandro Sánchez’ laboratory for help with transcriptome assembly and continuing improvements; Shaheena Bashir in Dr Gary Bader’s laboratory with assistance running baySeq analyses on our RNA-seq samples; and Chris Keay at High Speed Imaging (www.hsi.ca) for assistance with the high-speed imaging of live cilia.