Our previous work demonstrated a key role for soxB1-2 in the differentiation and function of sensory neuron subclasses in the planarian Schmidtea mediterranea (Ross et al., 2018). A subset of soxB1-2-regulated genes is abundantly expressed in a discrete pattern called the dorsal and peripheral ciliated stripes (Figure 1A), which contain ciliated sensory neurons involved in detecting water flow (rheosensation). We took a candidate-based approach to gain mechanistic insight into how the sensory stripe cells are specified from a heterogeneous soxB1−2⁺progenitor pool. POU4 genes are involved in the development of sensory organs detecting mechanical stimulation in divergent organisms (Manley and Ladher, 2008; Zhao et al., 2020). Thus, we investigated the expression and function of S. mediterranea pou4 genes. The planarian genome encodes two POU4 homologs, pou4-1 (also referred to as pou4-like) and pou4-2 (Onal et al., 2012). In previous work, we found that pou4-1 is expressed in the planarian CNS (Cowles et al., 2014). In contrast, analysis of pou4-2 using whole-mount in situ hybridization (WISH) showed expression localized in the dorsal head tip and dorsal and peripheral ciliated stripes of intact planarians (Figure 1A–B), a stereotyped pattern common to rheosensory genes (Ross et al., 2018). In addition, pou4-2 expression was detected in cells dispersed throughout the body in a subepidermal punctate pattern and in the cephalic ganglia and ventral nerve cords (Figure 1B). Because POU4 genes have been implicated as terminal selectors in diverse organisms (Leyva-Díaz et al., 2020), we examined the expression pattern of pou4-2 in regeneration blastemas to determine whether its activation coincides with late differentiation stages. During the first 24 hr of regeneration, pou4-2 expression was absent from the blastema. We first detected clear expression on day 3 of regeneration (Figure 1C). The patterning of pou4-2 expression in the blastema at day 3 was less organized and not confined to its normal spatial location, with pou4-2⁺ cells sparsely scattered throughout the regeneration blastema. During days 4 and 5, pou4-2⁺ cells began to repopulate the stereotypical stripe expression pattern, and by day 7, proper patterning was restored. In planarians, the regeneration blastema is populated by post-mitotic progenitors (Reddien, 2018). The delayed re-establishment of pou4-2 expression suggests that it is required in the later stages of cell differentiation in regeneration. Interestingly, embryonic expression of pou4-2 is predominantly detected during Stages 5–6, which coincide with organogenesis and nervous system development in S. mediterranea (Figure 1—figure supplement 1). While the functional role of pou4-2 in embryogenesis remains unknown, these data suggest that its temporal dynamics are consistent with a role in late-stage differentiation, as observed in adult regeneration (Davies et al., 2017).

soxB1-2 is expressed in and regulates transcription in dorsal and peripheral ciliated stripe neurons as well as in other neural populations, and the epidermis of S. mediterranea (King et al., 2024; Ross et al., 2018). Therefore, we searched the existing scRNA-seq data from the entire body and brain (Fincher et al., 2018) to examine the potential relationship between soxB1-2 and pou4-2. First, we extracted 1427 putative neuronal cells expressing soxB1-2. We resolved 19 distinct soxB1-2⁺ neuronal clusters (Supplementary file 1 and Figure 2A), of which cluster 8 was marked by pou4-2. The presence of synapsin and synaptogamin (neural markers) in cluster 8 indicated that these cells are neurons (Figure 2B). We combed through the dataset to identify genes that are differentially expressed in the pou4-2⁺ cluster (Figure 2C; Supplementary file 2); highly enriched genes included pkd1L-2 and hmcn-1-L, which are highly enriched in the planarian rheosensory organ (Ross et al., 2018). Because pkd1L-2 and hmcn-1-L expression requires soxB1-2 activity, we hypothesized that soxB1-2 regulates pou4-2 expression. Thus, we treated planarians with soxB1-2 or pou4-2 dsRNA (the RNAi treatment scheme is depicted in Figure 2—figure supplement 1A) and processed them for WISH. We observed a significant reduction in mechanosensory neuron-patterned pou4-2 expression in soxB1-2(RNAi) planarians, whereas pou4-2 expression in the central nervous system remained unaltered (Figure 2D). Conversely, soxB1-2 expression was downregulated in mechanosensory neuron-patterned areas important for rheosensation (Figure 2—figure supplement 1B). However, other areas enriched with soxB1-2 expression, such as the auricles - anteriorly positioned lateral flaps involved in chemotaxis (Almazan et al., 2021), the pharynx - an organ serving as the entrance and exit to the digestive system (Ishii, 1962), and the epidermis, were unaffected by pou4-2 RNAi. Together, these results support a model in which soxB1-2 positively regulates pou4-2 expression specifically in mechanosensory neurons, without affecting pou4-2 expression in other neural or epidermal populations.

In many organisms, pou4 and the proneural atonal genes are part of the same gene regulatory network (Leyva-Díaz et al., 2020). In Ciona intestinalis, atonal and pou4 are part of a regulatory cascade downstream of Notch signaling that generates sensory neurons (Joyce Tang et al., 2013). In mice, atoh1 is required for differentiation of multiple mechanosensory neuron types and stimulates expression of pou4f3 to promote hair cell fate (Yu et al., 2021), and overexpression of pou4f3 together with atoh1 and gfi1 in mouse embryonic stem cells can induce inner ear hair cell differentiation in vitro (Costa et al., 2015). There are three atonal homologs in the planarian genome, but none appear to operate in the same regulatory network as pou4-2. atoh-1 is expressed in a discrete neuronal population in the cephalic ganglia, while atoh8-1 and atoh8-2 are expressed in stem cells in the mesenchyme (Cowles et al., 2013). pou4-2 expression was unaffected after RNAi inhibition of all atonal genes in regenerated planarians (Figure 2—figure supplement 2). Thus, the functional relationship between pou4 and atonal observed in other animals does not appear to be conserved in planarians, based on current expression data and RNAi analyses.

Based on known roles of Pou4 genes, we hypothesized that pou4-2 is required for sensory neuron differentiation and function. To test our hypothesis, we performed RNAi of pou4-2 and pinpointed time points wherein the pou4-2 transcripts were robustly downregulated (not shown) and subsequently performed whole-animal RNA-seq on day 12 of the RNAi knockdown for pou4-2(RNAi) and control animals (see Materials and methods). Analysis of the resulting data revealed downregulation of putative pou4-2 target genes (Supplementary file 3). Because Pou4 genes are predicted to function as transcriptional activators, we focused further analyses on the downregulated gene set. pou4-2 RNAi RNA-seq uncovered 72 significantly downregulated genes (FC ≥1.4, adjusted p-value <0.1; Figure 3A and Supplementary file 3). GO analysis of the pou4-2-downregulated gene set revealed significant enrichment in ‘Mechanosensation’ (Supplementary file 4), including previously characterized genes we assessed to have roles in planarian mechanosensory modalities like vibration sensation and rheosensation, such as the polycystic kidney disease gene homologs pkd1L-2 and pkd2L-1 genes (Ross et al., 2024; Ross et al., 2018). This result is consistent with prior observations of the pou4-2 RNAi phenotype in scRNA-seq studies (King et al., 2024) and studies on the role of Notch signaling in planarians (Elliott, 2016). The discrete expression and pou4-2 RNA-seq dataset motivated us to characterize the regulatory role of this transcription factor in planarian sensory neuron function and regeneration.

The mechanosensory neurons in the rheosensory organ are distinguished by the expression of multiple sensory neural function genes and consist of at least two distinct populations, marked by the expression of terminal markers polycystic kidney disease 1 like-2 (pkd1L-2) and hemicentin-1-like (hmcn-1-L) (Ross et al., 2018). The evolutionarily conserved PKD1L-2 is a cation channel pore component required for mechanosensation and cilia function (Patel, 2015) pkd1L-2(RNAi) planarians exhibit prominent rheosensory defects (Ross et al., 2024; Ross et al., 2018). On the other hand, hmcn-1-L is an extracellular matrix component involved in anchoring mechanosensory neurons to the epidermis (Vogel and Hedgecock, 2001); no detectable sensory defects were observed in hmcn-1-L(RNAi) planarians (Ross et al., 2018). Consistent with the scRNA-seq data, we found that a subset of pou4-2⁺ cells co-expressed pkd1L-2 (74.8%) or hmcn-1-L (28.4%), representing two distinct mechanosensory neuron subtypes (Figure 3B, Supplementary file 7). We consistently observed variable expression levels; some cells showed high expression of pou4-2 and low expression of terminal markers (arrows in Figure 3B), while others showed lower expression of pou4-2 but high expression of terminal markers (arrowheads in Figure 3B), and some had high expression of pou4-2 and the terminal markers (white dashed box in Figure 3B). We also observed pou4-2⁺ cells lacking detectable expression of either pkd1L-2 or hmcn-1-L, which may represent late-stage progenitors that have not yet initiated terminal marker expression. Thus, the variable pou4-2 and terminal marker expression could be due to pou4-2 transcripts initially appearing at high levels in differentiating progenitor cells (initially terminal marker negative) to activate transcription of terminal markers and then persisting at lower levels in terminally differentiated cells.

Next, we asked if the function of pou4-2 is required to maintain and regenerate pkd1L-2 and hmcn-1-L expression in the dorsal and peripheral ciliated stripes. We conducted WISH on intact and regenerated planarians treated with dsRNA over a time course (Figure 1—figure supplement 1A). Like pou4-2, in control intact or regenerated planarians, pkd1L-2 and hmcn-1-L were expressed in the head tip, dorsal ciliated stripe, and dorsal and ventral peripheral stripes (controls in Figure 3C). In pou4-2(RNAi) animals, expression of both marker genes was strongly reduced, particularly in intact animals. In regenerating animals, expression was detected at lower levels and in scattered patterns. Interestingly, in the regenerates, minimal pkd1L-2 and hmcn-1-L expression was observed in the regeneration blastema. We also detected lower levels of expression of hmcn-1-L at scattered locations near peripheral stripes that were unaffected by pou4-2 RNAi (Figure 3C, dashed red boxes). These results indicate that pou4-2 is necessary for maintaining pkd1L-2 and hmcn-1-L expression in the most prominent dorsal and peripheral ciliated stripe cell populations. The persistence of hmcn-1-L expression in some peripheral cells may reflect partial knockdown efficiency or that pou4-2 function is not required in all hmcn-1-L⁺ cells.

To test whether the population of pou4-2⁺/terminal marker^- cells in dorsal ciliated stripe constitutes sensory neuron progenitors that have yet to express terminal markers in support of our hypothesis above, wild-type planarians were X-ray-treated with ~100 Gy, a dose reported to deplete early progenitors after 24 hr and late progenitors within 7 days (Eisenhoffer et al., 2008). We performed WISH analysis at 0, 3, 4, 4.5, 5, and 5.5 days post-irradiation (dpi). We quantified pou4-2⁺ cells lacking expression of pkd1L-2 and hmcn-1-L, compared to pou4-2⁺ cells co-labeled using a single fluorophore mix of pkd1L-2 and hmcn-1-L riboprobes (Figure 3—figure supplement 1A). By 5.5 dpi, we noticed an obvious decrease in pou4-2⁺/terminal marker^- compared to the number of pou4-2⁺/terminal marker⁺ cells (Figure 3—figure supplement 1B–C). The temporal decline in pou4-2⁺ cells closely resembled that of agat-1⁺ late progenitors, suggesting a similar position within the differentiation trajectory. In contrast, piwi-1⁺ neoblasts and prog-1⁺ early progenitors were almost entirely depleted by 2–3 days dpi, supporting the idea that pou4-2 marks a later-stage progenitor population (Figure 3—figure supplement 1D). These data support a model in which pou4-2 expression is established in late-stage progenitor cells prior to the expression of terminal sensory neuron genes and remains expressed at lower levels to maintain expression of sensory function genes in terminally differentiated mechanosensory neurons.

To further investigate the role of pou4-2 in regulating the differentiation of mechanosensory neurons, we selected eight additional genes from either the pou4-2 RNAi-downregulated gene set identified by RNA-seq or the S. mediterranea scRNA-seq database (Figures 2–3; see Supplementary file 2 for the list of genes). Four of the selected genes, cadherin-23 (cdh-23), Ephrin Receptor 1 (EphR1), lipoxygenase homology domain-1 (loxhd-1), and unconventional myosin VIIA (myo7a) are predicted to encode proteins homologous to those required for the proper function of inner ear hair cells in humans. Mutations in the human homologs of these genes are associated with sensorineural hearing loss (described below). We used WISH and RNAi analyses to determine their spatial expression patterns and assess whether expression was downregulated following pou4-2 or soxB1-2 knockdown (Figure 4A–B). The expression patterns of calmodulin-2 (calm-2), loxhd-1, and dd_28678 were confined to the stereotypical mechanosensory neuron pattern in the head tip, body periphery, and dorsal ciliated stripe, and were completely depleted in pou4-2(RNAi) and soxB1-2(RNAi) planarians (Figure 4A). pou4-2 RNAi-mediated loss of calm2 expression is consistent with observations reported by King et al., 2024. Calmodulins are important for ion channel activity and signal transduction, and human CALM2 mutations are associated with delayed neurodevelopment and epilepsy (Crotti et al., 2013). loxhd1 is predicted to encode a highly conserved stereociliary protein involved in hair cell function, and mutated human LOXHD1 causes DFNB77, a form of progressive hearing loss (Grillet et al., 2009). NOP2/Sun RNA methyltransferase family member 7 (nsun-7) was expressed in fewer cells in the mechanosensory neuron pattern overall, but an additional subset of nsun-7⁺ cells present in the optic cups was unaffected by pou4-2 and soxB1-2 inhibition, in contrast to the clearly reduced nsun-7 expression in sensory mechanosensory neuron-rich areas (Figure 4A). In humans, NSUN7 activity is required for proper flagella movement and sperm motility, and mutations result in male infertility (Khosronezhad et al., 2015). In control and pou4-2(RNAi) planarians, expression of the monocarboxylate transporter gene solute carrier family 16 member 24 (slc16a-24) was detected in the auricles. Auricular slc16a-24 expression was downregulated in soxB1-2(RNAi) but not pou4-2(RNAi) planarians, while expression in the mechanosensory neurons important for rheosensation was downregulated in both (Figure 4A).

Other genes we chose to analyze were not exclusively expressed in the ciliated stripes (Figure 4B). Expression of cadherin-23 (cdh23) was highest in the photoreceptors, and low expression was detected in the epidermis. While cdh23 expression was downregulated in the ciliated stripes, cdh23⁺ cells in the epidermis and photoreceptors were unchanged after pou4-2 and soxB1-2 inhibition. Human CDH23 is expressed in the sensory epithelium of the inner ear, where it is involved in maintaining the stereocilium organization of hair cells required for sound perception and equilibrioception (Kazmierczak et al., 2007), and CDH23 mutations are known to cause hereditary hearing loss (Woo et al., 2014). Smed-EphR1 (EphR1), encoding an ephrin receptor homolog, was also among the differentially expressed genes in soxB1-2⁺/pou4−2⁺neurons (Figure 2C). The role of Ephrin signaling in axon guidance is well-established and highly conserved; in mammals, the binding of ligand Efnb2 to receptor EphA4 is critical to the differentiation and patterning of hair and support cells on the cochlear sensory epithelium (Defourny et al., 2019) and for targeting and innervating auditory projections to hair cells (Defourny et al., 2013). In humans, mutations in EPHA4 and disruption of ephrin ligand binding are associated with sensorineural hearing loss (Lévy et al., 2018). While mechanosensory neuronal patterned expression of EphR1 was downregulated after pou4-2 and soxB1-2 inhibition, low expression in the brain branches of the ventral cephalic ganglia persisted (Figure 4B). EphR1 expression in the anterior-most region of the head tip was downregulated in soxB1-2(RNAi) but was unaffected in pou4-2(RNAi) planarians (Figure 4B). gelsolin-2 (glsn-2) was highly expressed in the sensory neuron pattern and in the epidermis, where it appeared most abundantly expressed near the body periphery and weakly expressed in the medial dorsal surface (Figure 5A). Gelsolins’ roles in nervous system development (Mazur et al., 2016) and as modulators of ciliogenesis and cilium length are evolutionarily conserved (Kim et al., 2010). Mechanosensory glsn-2 expression was downregulated in both pou4-2(RNAi) and soxB1-2(RNAi) planarians. Epidermal expression of glsn-2 was also reduced in soxB1-2(RNAi) animals, consistent with its broader role in ciliated epidermal cell maintenance (Ross et al., 2018).

In the dorsal ciliated stripe, body periphery, and head tip, low expression of unconventional myosin VIIA (myo7a) was detected and depleted in pou4-2(RNAi) and soxB1-2(RNAi) planarians (Figure 4B). Additionally, myo7a was highly expressed in the photoreceptors, and this expression remained in pou4-2(RNAi) and soxB1-2(RNAi) planarians. The low expression of myo7a detected beneath the epidermis was also unaffected following pou4-2 and soxB1-2 inhibition. Human MYO7A is important in stereocilium organization, differentiation, and signal transduction of inner ear hair cells (Jaijo et al., 2007). Defective MYO7A and CDH23, to a lesser extent, cause Usher Syndrome Type 1B (USH1B), which is characterized by deafness and reduced vestibular function (Roux et al., 2006). neuropeptide precursor-3 (npp-3) was highly expressed in the cephalic ganglia, ventral nerve cords, pharynx, and at lower levels in the parenchyma. A small subset of npp-3⁺ cells in the dorsal ciliated stripe was depleted in pou4-2(RNAi) and soxB1-2(RNAi) planarians (Figure 4B). Despite the presence of pou4-2⁺ cells in the ventral nerve cords (Figure 1B), pou4-2 inhibition appeared not to affect npp-3 expression in that region (Figure 5B).

Given that pou4-2 expression is reduced in soxB1-2(RNAi) planarians (Figure 2D) and that pou4-2(RNAi) animals show downregulation of mechanosensory and cilia-related genes (Figures 3–4), we reasoned that a subset of pou4-2⁺ cells represents terminally differentiated ciliated sensory neurons. Moreover, the requirement of soxB1-2 for maintaining ciliated epidermal and sensory neuron populations supports the idea that pou4-2 functions as a terminal selector in planarians. To assess the role of pou4-2 in the dorsal and peripheral ciliated stripes, we first immunostained control and RNAi-treated planarians with anti-Acetylated-Tubulin to mark cilia. Compared to the controls, pou4-2(RNAi) planarians showed a disruption of the stereotypical banded pattern with decreased cilia labeling along the dorsal ciliated stripe, while ciliated lawns on the dorsal and ventral surfaces remained unchanged (Figure 5A–B). To investigate the role of pou4-2 in mechanosensory function, we used a semi-automated behavioral assay to evaluate the mechanosensory response to vibration stimulation (Ross et al., 2024; Figure 5C). Knockdown of pou4-2 led to a significant reduction in vibration-induced contraction behavior in both intact and regenerating animals (Figure 5D–E). Thus, we conclude pou4-2 is downstream of soxB1-2 and is necessary for maintaining and regenerating ciliated mechanosensory neurons in the rheosensory organ (Figure 5F).

The interplay between lineage-specifying transcription factors and their respective gene regulatory networks coordinates precise developmental processes and stem cell fate decisions. POU4 transcription factors are conserved terminal selectors of sensory neuron fate, but the role of Pou4 has not been extensively characterized in regeneration. This study’s objective was to elucidate the role of Smed-pou4-2 (pou4-2) in planarian sensory neuron regeneration and to identify gene regulatory network components responsible for maintaining mechanosensory neuron function. pou4-2⁺ cells include ciliated mechanosensory neurons that allow planarians to detect water currents and vibrations (rheosensation), a function similar to POU4F3 in hair cells of the mammalian inner ear sensory epithelium responsible for auditory perception and equilibrioception. We showed that pou4-2 regulates the expression of genes homologous to those involved in stereocilium organization, cell adhesion, and nervous system development in other organisms. Several of these human homologs are implicated in sensorineural hearing loss. These findings shed light on the molecular mechanisms underlying planarian regeneration and provide insight into the conserved role of POU4 transcription factors in sensory neuron development across divergent species.

Differential expression analysis following pou4-2 knockdown revealed several genes homologous to those with known roles in mechanosensation in other organisms, including loxhd-1, cdh23, and myo7a. In vertebrate systems, mutations in these genes disrupt mechanosensory function and contribute to hearing loss. For example, a recent study has demonstrated that Loxhd-1 mutation in the inner hair cell does not affect the structural integrity of the hair cell bundle but rather prevents the activation of MET (mechanoelectrical transducer) channels, thereby contributing to progressive hearing loss in mice (Trouillet et al., 2021). Other interesting candidates included a planarian homolog of Ephrin receptors, EphR1, which was also expressed in soxB1-2⁺/pou4−2⁺neurons and was downstream of these two transcription factors. This observation highlights the potential for planarians to serve as a discovery platform for conserved regulators of sensory neuron patterning. We discovered that EphR1 is required for patterning of mechanosensory neurons in S. mediterranea (McCubbin, 2022; Warner, 2024), which led us to examine in detail how Ephrin signaling genes contribute to neural patterning in planarians (unpublished observations). Thus, as we demonstrated with soxB1-2 (Ross et al., 2024; Ross et al., 2018), planarians are also useful to analyze pou4-2-regulated genes and their roles in cell differentiation or mechanosensation. It will be important to perform a comparative analysis of Pou4-regulated genes gleaned from other animals, like sea anemones or vertebrates like birds and fish, which can regenerate hair cells or the lateral line, respectively (Chen et al., 2019; Tournière et al., 2020; Zhao et al., 2020). However, we found that our RNA-seq experimental design had limitations in detecting pou4-2-regulated transcripts, likely due to the use of systemic RNAi and bulk RNA extraction from whole animals. Before leveraging existing transcriptomic data on Pou4 homologs or hair cell or lateral line regeneration (Jiang et al., 2014; Ku et al., 2014; Tournière et al., 2020) for orthologous comparisons in other species to test to what extent the Pou4 gene regulatory network is conserved among these widely divergent animals, we have designed new experiments to enrich for pou4-2-expressing planarian tissues and have performed RNA-seq experiments producing a larger differentially expressed gene set (unpublished). In addition, ATAC-seq experiments could be performed to examine how pou4-2 activity affects chromatin architecture. These future genomic experiments should build upon this work and improve the resolution of the pou4-2 gene regulatory network implicated in sensory neuron regeneration.

pou4-2 function could encompass additional roles other than the ones identified in this study. In addition to the pkd1L-2⁺ and hmcn-1-L⁺ populations present in the rheosensory organ, there is a population of pou4-2⁺ cells in the ventral nerve cords and cephalic ganglia, which are not regulated by SoxB1-2 (Figure 1A). Although the ventral nerve cords are populated by npp-3⁺ cells, Pou4-2 activity does not regulate npp-3 expression in this region as it does in the rheosensory organ (Figure 4B), and it has not been confirmed whether any cells in the ventral nerve cords are pou4-2⁺/npp-3⁺ co-expressing cells. While this study did not define the function of this pou4-2⁺ central nervous system population, it may be possible to mine new scRNA-seq datasets to uncover transcription factors, such as in King et al., 2024, to predict the identities of pou4−2⁺ cells negative for expression of sensory neuron markers like pkd1L-2 and hmcn-1-L.

Our results indicate that the regulatory relationship between pou4 and atonal observed in other species does not appear to be conserved in planarians. The expression patterns of planarian atonal genes indicated that they represent completely different cell populations from pou4-2-regulated mechanosensory neurons. Although several pou4-2-regulated genes are also expressed outside the rheosensory organ, pou4-2 appears to primarily regulate gene expression in mechanosensory neuron-enriched regions, consistent with prior observations in pou4-2(RNAi) planarians (Elliott, 2016; King et al., 2024).

RNAi of pou4-2 substantially reduced the expression of the terminal mechanosensory markers pkd1L-2 and hmcn-1-L in intact animals, whereas pou4-2(RNAi) regenerate animals had a markedly reduced, dispersed expression of these genes at the head tips and peripheral ciliated stripes (Figure 3B). These findings raise the possibility that following injury, pou4-2⁺ progenitors or committed precursors may still undergo limited differentiation into sensory neurons in response to injury and polarity cues. This is consistent with previous observations in animals treated with hydroxyurea (HU) to block stem cell progression through S-phase. In these experiments, Smed-APC-1(RNAi) and Smed-ptc(RNAi) HU-treated animals were still able to regenerate neurons (Evans et al., 2011). It is also consistent with the hypothesis that many planarian stem cells are already specialized (Raz et al., 2021) and may have been unaffected by our RNAi treatment scheme. Nevertheless, both pou4-2(RNAi) intact animals and regenerates exhibited a significant reduction in mechanosensory responses compared to controls (Figure 5D–E). Moreover, the late re-expression of pou4-2 during regeneration, combined with the more rapid depletion of pou4-2⁺ cells lacking pkd1L-2 and hmcn-1-L expression compared to pou4-2⁺ cells co-expressing these terminal markers in irradiated animals, supports the idea that pou4-2⁺ progenitors give rise to terminally differentiated mechanosensory neurons (Figure 3—figure supplement 1).

pou4-2⁺ cells in the dorsal head tip and peripheral and dorsal ciliated stripes (the planarian rheosensory organ) are downstream of SoxB1-2 activity, and RNAi and WISH experiments demonstrated that pou4-2 expression is necessary for maintaining the functional properties of mechanosensory neurons (Figure 2D–E). However, the planarian rheosensory organ is composed of both ciliated mechanosensory neuronal and epidermal populations. Since soxB1-2 is a SoxB1 family transcription factor involved in ectodermal lineage specification, RNAi of soxB1-2 resulted in the loss of both ciliated sensory neuronal and epidermal populations in the rheosensory organ of planarian; thus, the dorsal ciliated stripe observed in acetylated tubulin labeling disappeared entirely (Ross et al., 2018). In contrast, in pou4-2(RNAi) animals, labeling of ciliated mechanosensory neurons was markedly reduced, but epidermal ciliated cells within the stripe appeared largely unaffected (Figure 5A–B). These findings support the conclusion that pou4-2⁺ cells represent a subset of mechanosensory neurons within the rheosensory organ and that pou4-2 is not required for the maintenance of ciliated epidermal cells, unlike soxB1-2, which has a broader role in ectodermal lineage regulation.

Despite molecular evidence indicating planarians possess ciliated mechanoreceptors sharing homology with mechanoreceptor function and development in other organisms, we have yet to fully resolve the cellular morphologies of the collection of cells comprising the ciliated stripes. Although it remains uncertain whether ciliated mechanoreceptors are products of convergent evolution or share a common cellular ancestry (Manley and Ladher, 2008), the role of Pou4 appears to represent a critical component of an adaptable gene regulatory network that has been co-opted to manufacture mechanoreceptors in distinct cell types. In addition, the mechanisms that specify pou4⁺ progenitors in response to local cues remain unclear. Studies have implicated Notch signaling as the likely culprit (Elliott, 2016), and recent studies in Schmidtea mediterranea have elegantly demonstrated that Notch signaling plays a role in patterning neurons and glial cells (Scimone et al., 2025). Together, our findings demonstrate that pou4-2 is critical for the maintenance, regeneration, and function of ciliated mechanosensory neurons in S. mediterranea. Future studies aimed at identifying additional pou4-2 target genes and defining how pou4-2 influences chromatin accessibility will provide deeper insight into the transcriptional regulation of mechanosensory regeneration in planarians.

Asexual clonal line CIW4 of S. mediterranea was maintained in 1 x Montjuïc salts (1.6 mM NaCl, 1.0 mM CaCl₂, 1.0 mM MgSO₄, 0.1 mM MgCl₂, 0.1 mM KCl, and 1.2 mM NaHCO₃) in the dark at 20 °C and fed weekly with pureed calf liver (Merryman et al., 2018). Planarians 3–5 mm in length were starved for 1 week before experimentation unless specified otherwise.

Sequences were obtained from an EST library (Zayas et al., 2005), cloned using gene-specific primers, or synthesized as eBlocks (IDT) and inserted into pPR-T4P (Liu et al., 2013) or pJC53.2 (Collins et al., 2010) vectors through ligation-independent cloning. Primer sequences, eBlock sequences, and EST clone accession numbers are listed in Supplementary file 5.

Riboprobes were synthesized using an in vitro transcription reaction from DNA templates with digoxigenin or fluorescein-labeled NTPs, and whole-mount in situ hybridizations were performed as previously described (King and Newmark, 2013) in an InsituPro automated liquid handling robot (CEM Corporation, Matthews, NC). Briefly, samples were incubated with anti-Digoxigenin-AP (1:2000, Roche) for chromogenic detection, and the signals were subsequently developed with NBT/BCIP in AP buffer. For double fluorescent in situ hybridizations (dFISH), samples were incubated for 16 hr at 4 °C with anti-DIG-AP and anti-FITC-POD (1:250, Roche). Peroxidase-conjugates were detected with tyramide signal amplification (TSA) as outlined previously in Brown and Pearson, 2015, and Fast Blue development was utilized for AP-driven reaction detection (Lauter et al., 2011).

To infer the gene expression profiles of pou4−2⁺ cells, we analyzed publicly available scRNA-seq data from S. mediterranea [GSE111764] (Fincher et al., 2018). From the whole-body data (50,562 cells) and the brain data (7,766 cells), we extracted putative neuronal cells, defined as those expressing at least one of the following transcripts: synapsin (dd_Smed_v4_3135_0_1), synaptotagmin (dd_Smed_v4_4222_0_1, dd_Smed_v4_6730_0_1, dd_Smed_v4_6920_0_1), and synaptosome associated protein 25 (dd_Smed_v4_13079_0_1, dd_Smed_v4_13255_0_1, or dd_Smed_v4_3977_0_1). This yielded 20,557 cells of which 1,427 expressed soxB1-2 (dd_Smed_v4_8104_0_1). Expression data were scaled and transformed using standard functions of the R library Seurat (Satija et al., 2015). We performed dimensionality reduction using UMAP and clustering using the Leiden algorithm (Traag et al., 2019). Since the UMAP projection suggested communities of disparate size, we used the Leiden algorithm in two steps – once with a smaller resolution to capture the large-scale structure (10 clusters) and then selectively with a higher resolution to resolve smaller communities, for a total of 19 distinct clusters. We characterized each cluster by finding differentially expressed genes (markers) using the bootstrap method of Pollard and van der Laan (Pollard and Laan, 2005) to calculate the z-score for every gene between cells in each cluster relative to cells outside the cluster (Supplementary file 1). The z-scores are then assessed for significance using the empirical Bayes method of Efron, 2008. The result is a posterior error probability lfdr assigned to each gene.

Bacterially expressed dsRNA was prepared by cloning gene-specific fragments into pPR-T4P or pJC.53.2 vectors, then transforming them into HT115 E. coli. Briefly, cultures grown overnight in LB with appropriate antibiotics were diluted in 40 mL 2xYT and incubated at 37 °C and shaking at 225 rpm. Once cultures reached 0.6–0.8 OD600, 1 mM IPTG was added to induce dsRNA synthesis, and incubation continued for 2 hr. Cultures were then pelleted at 3,000 x gg for 10 min at 4 °C, resuspended in 8 mL LB, and aliquoted into eight microcentrifuge tubes. Resuspended cultures were pelleted at 11,000 x gg for 5 min at 4 °C, aspirated, and stored at –80 °C. For RNAi feeding, bacterial pellets were mixed with liver puree as previously described (Gurley et al., 2008). Animals were fed eight times over 4 weeks, with gfp dsRNA used as a negative control. For stainings, intact animals were fixed 10 days after the 8th feed, and regenerated animals were pre-pharyngeally amputated one day after the 8th feed and fixed after 10 days of regeneration. RNAi feedings for the behavioral assays were performed using in vitro transcribed dsRNA mixed with pureed liver and agarose, as described in Ross et al., 2024. For intact animals, assays were performed three days after the 8th feed. For regenerates, animals were amputated 24 hr after the 8th feed and tested for behavioral defects 14–17 days of regeneration. The total numbers of animals for RNAi experiments are summarized in Supplementary file 6.

Three biological replicates were obtained, each consisting of four worms of approximately four mm length at the start of the experiment, which were starved for one week prior to the start of RNAi feeding. Worms were fed bacterially expressed dsRNA three times on days 0, 3, and 7, and RNA was extracted and purified on day 12 (Allen et al., 2021). 500 ng of total RNA was used for the RNA-seq library preparation and sequencing at MedGenome, Inc (Foster City, CA). The Poly-A-containing mRNA molecules were purified using poly-T oligo attached magnetic beads, and then the mRNA was converted to cDNA using Illumina TruSeq stranded mRNA kit (20020595) according to the manufacturer’s protocol. Libraries were sequenced for 100 cycles to a depth of 30 million paired reads using Illumina NovaSeq 6000 (Illumina, San Diego, CA). The following quality control steps were performed on the fastq files: Base quality score distribution, Sequence quality score distribution, Average base content per read, GC distribution in the reads, distribution of over-represented sequences, and adapter trimming. Based on the quality report of fastq files, sequences were trimmed wherever necessary to retain only high-quality sequences for further analysis. In addition, the low-quality sequence reads are excluded from the analysis. Data quality check was performed using FastQC (v0.11.8). The adapter trimming was performed using the fastq-mcf program (v1.05) and cutadapt (v2.5; Martin, 2011). Transcriptome alignment was performed using RSEM (version RSEM v1.3.1; Li and Dewey, 2011) against the dd_Smed_v6 transcriptome (Rozanski et al., 2019) to build Bowtie transcriptome indexes using rsem-prepare-reference, then used rsem-calculate-expression for aligning and expression calculation. Differential expression analysis was performed using DESeq2 (R Bioconductor package; Anders and Huber, 2010) with default parameters, and then differentially reduced genes were defined as those having a fold-change of <1.4 and p-adjusted value <0.1 (see Supplementary file 3). The RNA sequencing data have been deposited in NCBI under BioProject accession PRJNA1258257. For Gene Ontology (GO) annotation and over-presentation analysis, the unique set of pou4-2(RNAi) downregulated transcripts was compared to the human proteome (BLASTX against the Swiss-Prot Homo sapiens proteome, cutoff e-value<1e^–3). Human UniProt IDs were used for enrichment analysis using Fisher’s Exact tests with FDR multiple test correction (FDR <0.05) in http://geneontology.org/. GO results are reported in Supplementary file 4.

Animals were sacrificed in ice-cold 2% HCl for 30 s, followed by incubation in Carnoy’s fixative (6 parts ethanol: 3 parts CHCl3: 1 part glacial acetic acid) for 2 hr at 4 °C (Forsthoefel et al., 2018), followed by a dehydration step with 100% methanol for 1 hr at 4 °C. The animals were bleached overnight under a lamp with 6% H₂O₂ in methanol and then rehydrated in 75%, 50%, and 25% methanol-PBSTx, followed by two 5-min PBSTx washes. PBSTb (1% BSA in PBSTx) was used for blocking at room temperature for 2 hr. Primary antibody labeling was carried out with mouse anti-Acetylated Tubulin (Sigma-Aldrich, St. Louis, MO) diluted in PBSTb (1:1000) overnight at 4 °C. Six 1 hr PBSTx washes followed by 1 hr of PBSTb blocking were performed before anti-mouse-HRP (1:1000, Cell Signaling) incubation overnight at 4 °C. After six 1 hr PBSTx washes, acetylated tubulin was detected through TSA development as described in Brown and Pearson, 2015 with the following exceptions: no 4-IPBA or dextran sulfate was added to the TSA reaction buffer, Cy3-tyramide was diluted 1:250, and development took place for a total of 20 min.

Analysis of the planarian’s ability to detect a vibration stimulus was conducted essentially as described in Ross et al., 2024. Briefly, groups of five control gfp(RNAi) or pou4-2(RNAi) planarians were added to a 100×15 mm petri dish containing 40 ml of 1 x Montjuïc salts that was placed inside a dish lid that was mounted to a cold LED lighted board using clear silicone paste and observed until gliding normally. Then, an Arduino-controlled arm delivered five taps at a rate of one tap every 75 ms to the side of the dish. Experimental runs were recorded on a Basler Ace 2 Pro ac1440-220uc camera connected to a PC running Basler’s pylon Viewer 64-bit version 6.3.0 software at a frame rate of 10 frames/sec and a frame size of 1440×1080 pixels. The video frames were analyzed in Fiji (ImageJ2 version 2.9.0; Schindelin et al., 2012), using the line tool to measure the longest pre-stimulus gliding length and the length of the worms following the stimulus. The percent change in length was calculated as [(Length_Prestimulus – Length_Poststimulus)/Length_Prestimulus]×100. Statistical analysis and graph generation were performed in GraphPad Prism 9 (GraphPad Software, Boston, MA). One-way ANOVA analyses were performed and corrected using Dunnett’s correction. All means were compared to the control group, and statistical significance was accepted at values of p<0.01.

Starved animals (3–4 mm) were irradiated with 100 Gy of X-rays (130 kV, 5 mA, 8.4 Gy/min) for approximately 12 min using a CellRad irradiator (Precision X-Ray, Madison, CT).

Brightfield images were acquired with a Leica DFC450 camera and M205 stereomicroscope. Animals processed for fluorescent in situ hybridization and immunohistochemistry were mounted in Vectashield diluted 1:1 in 80% glycerol. Fluorescent images were acquired using a Zeiss AxioZoom equipped with an Apotome using Zen Pro version. High-magnification acetylated tubulin images were acquired using a Zeiss Axio Observer Inverted Microscope equipped with an AV4 Mod Apotome using AxioVision v4.6 (Carl Zeiss Microscopy, LLC, White Plains, NY).

For co-labeling experiments between pou4-2-regulated genes with pkd1L-2 or hmcn-1-L, maximum intensity projections of stacked fluorescent images were acquired with a depth of 12 µm. For pou4-2 co-labeling with pkd1L-2 and hmcn-1-L in irradiated planarians, maximum intensity projections of stacked fluorescent images were acquired with a depth of 24 µm. The regions quantified spanned the width of the sensory neuron expression pattern in the head tip and 500 µm anterior to the head tip along the length of the dorsal ciliated stripe. Cells were manually counted using Zeiss ZEN lite v3.3. Cell quantification was represented as the number of positive cells per mm², and graphs were made using GraphPad Prism (GraphPad Software, Boston, MA). Three to six biological replicates per group were used to quantify co-labeling. Cell quantification details and the total number of cells counted for Figure 3B, Figure 3—figure supplement 1A–B are summarized in Supplementary files 7 and 8, respectively.