aristaless-like homeobox-3 is wound induced and promotes a low-Wnt environment required for planarian head regeneration

Adult multi-tissue regeneration requires the simultaneous integration of wound signals, proliferation and patterning of new tissues as they are made (Tanaka and Reddien, 2011). Tissue-resident adult stem cells (ASCs) often have the capacity to respond to wound signals and drive the compensatory proliferation that is required to remake missing tissues (Pearson and Alvarado, 2008). In addition, signaling pathways that function in specifying position and polarity in the developing embryo are often reactivated during regeneration (Gurley et al., 2010). Decades of work have gone into describing the cascades of wound responses, proliferative pathways of ASCs, and how patterning is required during various regenerative contexts, yet it is not always clear how these processes are integrated to ensure successful regeneration. In order to understand the mechanisms of adult tissue regeneration, a model system with this biological capability can yield valuable insights.

The asexual freshwater planarian Schmidtea mediterranea (S. med.) is a constitutive adult, capable of regenerating all organs after injury (Reddien and Alvarado, 2004). The remarkable regenerative abilities of planarians are reliant on a large population of cycling ASCs – called neoblasts – that constitute ∼20% of all somatic cells, some of which are pluripotent (Aboobaker, 2011; Wagner et al., 2011). Neoblasts are the only known dividing cells, are broadly distributed throughout the parenchyma and express the piwi homolog smedwi-1 (hereafter referred to as piwi-1) (Reddien et al., 2005). Although neoblasts are molecularly heterogeneous, they are also extremely plastic in their cell fates, e.g. posteriorly located neoblasts in the tail can remake all anterior structures after amputation (Scimone et al., 2014; van Wolfswinkel et al., 2014). Thus, immense effort has been put forth on understanding the spatial patterning of both new and existing tissues during the regeneration process.

In planarians, positional control genes (PCGs) are expressed in muscle cells where they control regional identity and are required for regeneration (Witchley et al., 2013). Perhaps the most important PCG pathway in planarian is Wnt/β-catenin signaling along the anterior-posterior (AP) axis, which is required not only to maintain tissue identity as cells turnover during homeostasis but it is also the key pathway for making the polarity decision of whether a wound is anterior or posterior facing (Reddien, 2011; Sureda-Gómez et al., 2016). At homeostasis, wnt1 is expressed at the posterior midline (posterior pole), whereas notum, a secreted Wnt inhibitor, is expressed at the anterior midline (anterior pole) (Petersen and Reddien, 2008, 2009a, 2011). Upon transverse injury, rapid changes in gene expression in stem cells and differentiated cells initiate regeneration (Wenemoser et al., 2012). Within 6-24 h post-amputation (HPA), the Wnt ligand wnt1 is expressed in subepidermal cells at both anterior-facing and posterior-facing wound edges (Petersen and Reddien, 2009b), whereas notum is asymmetrically expressed at anterior-facing wounds (Petersen and Reddien, 2011). Sustained wnt1 expression at posterior-facing wounds promotes tail regeneration, whereas the notum-dependent low-Wnt environment at anterior-facing wounds stimulates head regeneration. After wound-induced notum expression, a stem cell-derived population of notum⁺ muscle cells localizes at the anterior midline by ∼72 HPA and anterior polarity is restored (Scimone et al., 2017; Witchley et al., 2013). Disruption of Wnt signaling by RNAi of β-catenin can transform injury sites into ectopic heads, whereas increasing Wnt signaling with RNAi of the negative Wnt regulator APC will transform injury sites into ectopic tails (Gurley et al., 2008; Petersen and Reddien, 2008).

Parallel to wound patterning by PCGs, injury causes specific cascades of wound-induced gene expression, and in turn, stem cells exhibit a well characterized mitotic response to amputation (Wenemoser and Reddien, 2010; Wenemoser et al., 2012; Wurtzel et al., 2015). Interestingly, stem cells throughout the body express Wnt signal transduction effectors, including all nine planarian frizzled receptors (Currie et al., 2016; Molinaro and Pearson, 2016; Wurtzel et al., 2015). The sum total of these wound programs drives neoblasts to remake new tissue of the appropriate cell types. However, this raises an intriguing question: how are wound-induced programs, positional identity programs and neoblast proliferation programs integrated to ensure proper regeneration of specific structures?

In the current study, we report a role for the aristaless-like homeobox-3 transcription factor, alx-3, in regulating anterior regeneration through the activation of wound-induced gene expression programs and the stem cell proliferative response. alx-3 was initially identified as a transcription factor enriched in a population of putative neural-fated neoblasts at homeostasis (Molinaro and Pearson, 2016). We find that, in uninjured animals, alx-3 is necessary for the maintenance of specific neuronal subtypes, including dopaminergic and GABAergic neurons, as well as for regulating brain size in proportion to body size. Surprisingly, we find that alx-3 expression is injury induced in neoblasts and muscle cells at anterior-facing wounds by 24 HPA. Animals where alx-3 has been knocked down by RNAi exhibit loss of anterior blastema formation and head regeneration in tail fragments. Interestingly, under conditions where the AP axis is not disrupted, alx-3(RNAi) animals are able to regenerate anterior tissues laterally. To explain the headless phenotype in alx-3(RNAi) tails, we find that alx-3 is required for anterior notum expression after 6 HPA and alx-3 expression itself is negatively regulated by Wnt signaling in a feedback loop. We conclude that alx-3 plays an essential role in integrating wound signals and executing specific responses from anterior-localized stem cells, which subsequently primes the region for neurogenesis and successful head regeneration.

The aristaless-like homeobox transcription factor alx-3 was initially identified in a single-cell transcriptomics study aimed at determining genes enriched in a putative population of neural-fated stem/progenitor cells in the homeostatic planarian head (Molinaro and Pearson, 2016). In this study, 96 single stem cells were isolated from the head region of the animal and were transcriptionally profiled. Interestingly, a gene signature was found that had stem cell gene expression, but also expressed known neural genes, which we termed nu (ν)-neoblasts (Molinaro and Pearson, 2016). From this, we identified six transcription factors enriched in the population of ν-neoblasts, of which alx-3 was a marker, but not yet functionally examined.

We first determined the wild-type homeostatic expression of alx-3 using whole-mount in situ hybridization and observed that alx-3 displayed strong expression in the brain along with dispersed body-wide expression (Molinaro and Pearson, 2016) (Fig. 1A). Published atlases of single-cell RNA-sequencing (scRNAseq) data reveal that alx-3 is detected in three main cell lineages: a subset of piwi-1⁺ stem cells; most neural cell clusters; and muscle cells (Fincher et al., 2018; Plass et al., 2018) (Fig. 1B). Further analysis of the alx-3 and piwi-1 co-expressing cells showed that they were annotated as ‘neural’ (Neural cluster 1), suggesting that this may be the ν-neoblast neural progenitor population (Fig. 1C). Previous bulk RNAseq data from cell populations on a flow cytometer showed expression for alx-3 was enriched in the ‘X2’ gate, which is associated with a mix of differentiating progenitor cells and G0/G1 neoblasts (Fig. 1D) (Hayashi et al., 2006; Labbé et al., 2012; Zhu et al., 2015). Double fluorescence in situ hybridization of alx-3 and piwi-1 validated co-expression of alx-3 in stem cells adjacent to the brain lobes at homeostasis (11±6.8% of piwi-1+ cells had detectable alx-3) (Fig. 1E).

To characterize the differentiated neuronal cell types that co-expressed alx-3, double fluorescence in situ hybridization (dFISH) was performed using neuronal subtype markers at homeostasis. Cholinergic neurons, labeled by the expression of the gene choline acetyltransferase (chat), are abundant in the planarian brain and a substantial proportion of cholinergic neurons expressed alx-3 (56±8.2%) (Fig. 1F). Dopaminergic neurons marked by the expression of the gene tyrosine hydroxylase (th) are located medially with respect to the brain lobes and a smaller proportion of dopaminergic neurons were found to co-express alx-3 (23±11%). Octopaminergic neurons, marked by the expression of the gene tyrosine beta-hydroxylase (tbh) are located within the medial region of the brain and only a small proportion of octopaminergic neurons co-expressed alx-3 (9.2±5.2%). GABAergic neurons, labeled by the expression of glutamic acid decarboxylase (gad), are present in two spatially distinct populations: one located more ventro-medially (VM) and the other located dorso-laterally (DL) (Nishimura et al., 2008). Within the VM subgroup, virtually all GABAergic neurons co-expressed alx-3 (97±3.0%), whereas roughly half of the DL subgroup co-expressed alx-3 (47±12%). As many cholinergic neurons are also located in the peripheral nervous system, we determined that peripheral alx-3⁺ cells also co-expressed chat and overlapped with immunostaining of the axonal marker 1H6 (Fig. S1) (Ross et al., 2015). In total, the expression data demonstrate that alx-3 is expressed in heterogenous population of neural subtypes in the brain, as well as a subset of piwi-1⁺ cells.

To determine the functional role of alx-3, we assessed changes to brain morphology and neuronal populations in uninjured animals. The effect of alx-3 knockdown on brain size and morphology was determined using DAPI staining and measuring brain length with normalization to the length of each individual animal. Although alx-3(RNAi) animals were generally larger in size after nine RNAi feeds, the ratio of brain length to body length was smaller compared with control animals, indicating a significant reduction in overall brain size (Fig. 2A). To determine whether reduced brain size was a result of dysregulated cell proliferation, immunostaining with anti-phosphohistone H3 (H3P) was performed in uninjured animals, labelling cells in the G2/M phase of the cell cycle. H3P staining showed no significant difference in cell proliferation levels between the control and alx-3 knockdown animals (Fig. 2B). To further support this finding, homeostatic worms were fed the thymidine analogue bromodeoxyuridine (BrdU) and fixed after a 7-day chase period. To determine whether BrdU was actively incorporated into new cells in the brain lobes, BrdU⁺ cells were quantified within the region of the brain lobes defined by the expression of the neural marker chat, revealing no significant difference in BrdU incorporation between alx-3 knockdown and control animals during the 7 days examined (Fig. 2C).

As alx-3 was expressed in several neuronal populations at homeostasis, expression of each neuronal marker was assessed under alx-3 knockdown conditions without injury. There was no substantial effect on the proportion of chat⁺ or tbh⁺ neuronal populations when alx-3 expression was inhibited (Fig. 2D,E). However, alx-3(RNAi) animals had significantly fewer dopaminergic neurons labelled by the expression of th, as well as fewer GABAergic neurons labelled by the expression of gad, observed in both VM and DL subpopulations (Fig. 2F,G). Together, these results reveal that although alx-3 is not required to maintain overall cell proliferation and overall neural differentiation at homeostasis, it is required to maintain appropriate brain size and number of gad⁺ and th⁺ neurons in uninjured animals.

To determine the expression pattern of alx-3 during regeneration, a regeneration assay was performed in which animals were amputated anterior and posterior to the pharynx, and each remaining fragment was allowed to regenerate for 0.5, 6, 24, 48 or 72 h. Whole-mount in situ hybridization of alx-3 revealed that alx-3 was detected by 24 HPA with much higher expression at anterior-facing wounds (Fig. 3A). To assess whether wound-induced alx-3 expression was in stem cells, FISH of alx-3 and immunostaining of PIWI-1 (using anti-PIWI-1 antibody, a kind gift from Dr Jochen Rink, Max Planck Institute for Multidisciplinary Sciences, Germany) was performed. Interestingly, alx-3 was detected in PIWI-1⁺ cells near the anterior wound edge (Fig. 3B). To further support this finding, an irradiation assay was performed to ablate stem cells. Whole animals were lethally irradiated and amputated posterior to the pharynx at 48 h post-irradiation, tail fragments were then stained for alx-3 expression at 48 HPA. This assay revealed that wound-induced alx-3 expression was also ablated in the absence of stem cells (Fig. 3C). Together, these results demonstrate that alx-3 expression is induced by injury in stem cells, primarily at anterior-facing wounds.

alx-3 expression is induced in stem cells after injury, thus we next explored the role of alx-3 during regeneration. After alx-3 RNAi, animals were amputated anterior and posterior to the pharynx, and each fragment allowed to regenerate for 14 days. Knockdown of alx-3 expression produced a noticeable phenotype where tail fragments fail to regenerate an anterior blastema and head (44% of tail fragments) (Fig. 3D). Trunk fragments, however, were consistently capable of head regeneration (Fig. S2), which is reminiscent of animals with knockdown of follistatin (Tewari et al., 2018). Owing to the similarity in the alx-3(RNAi) and follistatin(RNAi) phenotype, we assessed whether follistatin expression was impacted by alx-3(RNAi). Although we observed a significant reduction in the number of follistatin⁺ cells at 48 HPA in alx-3(RNAi) tail fragments, overall follistatin expression was largely intact (Fig. S3).

alx-3(RNAi) head and trunk fragments did not exhibit regeneration defects compared with tail fragments, despite similar levels of knockdown (Fig. S4), so we then focused on how head regeneration was disrupted in tail fragments only. Compared with controls, alx-3(RNAi) tail fragments that exhibited loss of head regeneration also resulted in loss of brain regeneration, as indicated by the disruption of chat and gpas fluorescence in situ hybridization labelling the brain lobes (Fig. 3E). alx-3(RNAi) tail fragments exhibited ventral nerve cords fusing at the anterior midline without regenerating brain tissue proper. As alx-3(RNAi) tail fragments do not regenerate a head, we wondered whether they were capable of regenerating other missing tissues. To assess this, we performed whole-mount in situ hybridization of the pharynx marker foxA (Adler et al., 2014), which revealed that when alx-3(RNAi) tail fragments do not regenerate a head; they also do not regenerate a pharynx (Fig. S5).

Failure in regeneration is often associated with abnormal stem cell dynamics, such as defects in cell proliferation or differentiation. To determine whether the loss of anterior regeneration may be due to dysregulated stem cell proliferation in alx-3(RNAi) tail fragments, cell proliferation was quantified using H3P. Planarians have two distinct waves of cell proliferation after injury, the first occurs at 6 HPA followed by a second peak at 72 HPA (Baguñà, 1976; Saló and Baguñà, 1984; Wenemoser and Reddien, 2010). Stem cell proliferation was significantly reduced during both mitotic peaks in alx-3(RNAi) animals (Fig. 3F). H3P staining of sagittally amputated alx-3(RNAi) animals did not result in a significant difference in stem cell proliferation during either mitotic peak. Similarly, the first mitotic peak in alx-3(RNAi) animals with incision injuries did not result in reduced stem cell proliferation (Fig. S6). As knockdown of alx-3 in uninjured animals did not show an appreciable difference in cell proliferation, we conclude that the role of alx-3 in injury-induced stem cell proliferation is specific to regeneration when the anteroposterior axis is disrupted.

To determine whether failure of brain regeneration was a result of defects in neurogenesis in the absence of alx-3, the well-established β-catenin(RNAi) phenotype was used as it results in the formation of ectopic heads and brain tissue at posterior regions of the animal, even without injury (Gurley et al., 2008, 2010; Petersen and Reddien, 2008, 2009b; Reuter et al., 2015; Witchley et al., 2013). Animals were given a combination of either control(RNAi) or alx-3(RNAi) for nine feeds, followed by two feeds of either control(RNAi) or β-catenin(RNAi) (Fig. 4A). After the RNAi knockdown regimen, the formation of ectopic brain tissue was assayed using fluorescence in situ hybridization of the neural markers gpas and pc2. In β-catenin and alx-3 double RNAi conditions, animals were able to generate ectopic neural tissue at posterior regions of the animal, similar to β-catenin(RNAi) animals (Fig. 4B). To investigate this result further, brain regeneration after minor injuries was evaluated. An incision or wedge cut-out was made below the right eye of animals without exceeding the midline and regeneration was assessed 7 days post-injury (DPI). alx-3(RNAi) animals were able to regenerate brain tissue following either injury type, similar to control animals (Fig. 4C).

Brain regeneration is often contingent on anterior polarity specification after injury, thus performing sagittal amputations, which do not require re-establishment of AP polarity, allows the separation of AP position from brain regeneration. In sagittally amputated animals, fluorescence in situ hybridization of pc2 revealed that the missing brain lobe in alx-3(RNAi) animals regenerated similar to control animals by 14 DPA (Fig. 4D). Fluorescence in situ hybridization of the anterior pole marker notum was performed to confirm that the anterior pole and overall AP polarity remained intact following amputation. We conclude from these experiments that alx-3 is not required for neural regeneration and that defects in neurogenesis do not explain the headless phenotype seen in alx-3(RNAi) tail fragments.

Planarian regeneration involves several tightly regulated transcriptional wound responses. The earliest wound response to occur in planarians is the transcription of early acting genes in differentiated epidermal and subepidermal cells within 30 min after injury (Wenemoser et al., 2012). These genes are rapidly transcribed as an immediate genomic response to injury and their roles may include wound closure, apoptosis, activation of downstream wound responses and stem cell proliferation (Wenemoser et al., 2012). We reasoned that dysregulated expression of genes involved in characterized wound responses in the absence of alx-3 expression would provide insight into the functional role of alx-3. To this end, we performed RNAseq of alx-3(RNAi) or control(RNAi) whole-tail fragments at 6, 24 and 72 HPA, and compared differential gene expression between the conditions. (Fig. 5A). Analysis of RNAseq data between control(RNAi) and alx-3(RNAi) tail fragments identified several dysregulated wound response genes (Fig. 5B). For example, fos-1, jun-1, plasminogen-1 and hadrian expression were downregulated at 6 HPA, whereas delta-1 and ston expression were upregulated at 6 HPA in alx-3(RNAi) tail fragments. The Runx-family transcription factor runt-1 is known to be wound-induced in stem cells and is required for daughter cell specification during regeneration (Wenemoser et al., 2012). We observed that in alx-3(RNAi) tail fragments, runt-1 was downregulated at each time point. Whole-mount in situ hybridization staining in tail fragments between control and alx-3 RNAi conditions supported the changes in gene expression observed in the RNAseq data (Fig. 5C). Ultimately, these results indicate that alx-3 is required to activate key components of the early and late transcriptional wound responses during head regeneration in tail fragments.

To determine whether the expression of early wound-response genes is required for wound-induced alx-3 expression, knockdown of fos-1 was performed. Similar to previous reports by Wenemoser et al. (2012), fos-1(RNAi) animals produced smaller blastemas, although no other severe regeneration defects were observed at 7 DPA (Fig. S7A). fos-1(RNAi) resulted in slightly reduced wound-induced notum expression at 24 HPA as well as reduced alx-3 expression at 48 HPA, although it was not significant enough to ablate anterior regeneration. (Fig. S7B and S7C, respectively.)

The Wnt signaling pathway plays a key role in the choice between anterior or posterior regeneration. Even in planarian species that cannot regenerate heads from tail fragments, if Wnt signaling is decreased, head regeneration can be restored (Liu et al., 2013; Sikes and Newmark, 2013; Umesono et al., 2013). The transcriptional injury response involves regulation of Wnt signaling genes at wound sites between 6 and 24 h post-injury (Gurley et al., 2010; Petersen and Reddien, 2009a, 2011; Wurtzel et al., 2015). Although there are many components, wnt1 and notum expression direct the regeneration outcomes depending on the context of the wound. (Tewari et al., 2018; Wenemoser et al., 2012; Wurtzel et al., 2015). For example, wnt1 is expressed at both anterior and posterior-facing wounds, whereas anterior-facing wounds asymmetrically express the Wnt inhibitor notum in order to achieve a low-Wnt environment required for head regeneration (Gurley et al., 2010; Petersen and Reddien, 2009a, 2011; Wenemoser et al., 2012).

Consistent with dysregulation of early-acting wound response genes, RNAseq data revealed downregulation of wnt1 and notum expression at all time points examined (Fig. 5B). To investigate the requirement of alx-3 during anterior regeneration polarity, alx-3(RNAi) tail fragments were stained for wnt1 and notum expression. A significant reduction in the number of wnt1-expressing cells at anterior-facing wounds was observed in alx-3(RNAi) tail fragments compared with the control(RNAi) at 6 HPA (Fig. 6A). Interestingly, the reduction in wnt1 expression was not observed at either anterior or posterior-facing wounds of trunk fragments (Fig. S8A). Although wnt1 is expressed broadly at both anterior- and posterior-facing wounds, determining the effect of alx-3 knockdown on notum expression at anterior-facing wounds was of particular importance considering the defect in anterior regeneration in alx-3(RNAi) animals. Whole-mount in situ hybridization of notum revealed a significant reduction in the number of notum-expressing cells compared with control(RNAi) tail fragments at 24 HPA (Fig. 6B). However, notum expression was also significantly reduced in trunk fragments at the same time point (Fig. S8B). These data demonstrate that alx-3 is required for activation of wnt1 and notum expression at anterior-facing wounds of tail fragments.

At 72 HPA of head regeneration, notum expression resolves into an anterior pole in a stem cell-dependent manner (Oderberg et al., 2017; Vogg et al., 2014). However, alx-3(RNAi) tail fragments did not regenerate the notum⁺ anterior pole at 72 HPA (Fig. 6C). To determine whether anterior pole formation was delayed in alx-3(RNAi) tail fragments, notum expression was assessed at 10 DPA; similarly, absence of notum expression was observed (Fig. 6C). Amputated trunk fragments were able to recover their notum-pole expression at 10 DPA (Fig. S8C). Notably, alx-3 was not required to maintain the population of notum⁺ cells in uninjured animals (Fig. 6D). These data demonstrate that alx-3 is required to regulate wnt1 and notum expression during head regeneration in tail fragments.

After amputation, positional information across the remaining tissue must rescale and reform expression domains to regenerate a correctly proportioned animal (Reddien and Alvarado, 2004). For example, the Wnt ligand wntP-2 is expressed in a posterior-to-anterior gradient and upon amputation; tail fragments initially express wntP-2 across the entire length of the fragment and will then rescale expression over the first week of regeneration (Gurley et al., 2010). To test whether overall body-plan reorganization was affected in alx-3 knockdown animals, whole-mount in situ hybridization of wntP-2 was performed. Unlike control animals that gradually repress wntP-2 expression from the anterior region of tail fragments between 48 HPA and 10 DPA, the expression of wntP-2 was observed at anterior-most regions of alx-3(RNAi) animals and was never properly rescaled (Fig. 6E). Taken together, these data indicate that alx-3 expression is required to regenerate anterior polarity and restore the AP axis in tail fragments.

Although alx-3 was wound induced in stem cells, we also observed alx-3 expression in notum⁺ and sfrp-1⁺ pole cells in uninjured animals (22±3.2% of notum⁺ cells, 21±6.8% of sfrp-1⁺; Fig. 7A). Thus, we reasoned that notum expression during head regeneration may require alx-3 cell-autonomously. We determined the dynamics of co-expression between alx-3 and notum during regeneration and found that alx-3 expression was observed in wound-induced notum⁺ cells at both 24 HPA (23±3.5%) and at 72 HPA (29±2.7%; Fig. 7B). Based on the phenotype and co-expression, we conclude that alx-3 is most likely to be required cell-autonomously in regenerating notum⁺ pole precursor cells.

Muscle cells play an important role in the expression of PCGs, thus we assessed the role of alx-3 in maintaining the muscle cell population. alx-3 is expressed in a subset of collagen⁺ muscle cells (11±2.3%) at anterior-facing wound sites at 24 HPA (Fig. S9A), and we assessed for any changes in body wall musculature in alx-3(RNAi) animals during regeneration. Fluorescence in situ hybridization for collagen revealed no significant difference in the proportion of collagen⁺ cells in alx-3(RNAi) tail fragments at 72 HPA or in intact animals compared with control(RNAi) (Fig. S9B and S9C, respectively). Similarly, there was no substantial difference in muscle fiber morphology or density in alx-3(RNAi) tail fragments at 72 HPA or in intact animals compared with control(RNAi), observed using 6G10 immunostaining labelling muscle fibers (Fig. S9D and S9E, respectively). Together, these negative data indicate that although alx-3 is expressed in a subset of muscle cells and alx-3(RNAi) results in dysregulated PCG expression occurring in muscle cells, this is not due to a reduction in the muscle cell population or to disrupted muscle fiber morphology.

To investigate whether Wnt signaling feeds back on wound-induced alx-3 expression, knockdown of β-catenin, APC or notum were performed followed by post-pharyngeal amputation and whole-mount in situ hybridization of alx-3 to assess wound-induced alx-3 expression 48 HPA (Fig. 7C). Knockdown of β-catenin inhibits Wnt signaling, which resulted in higher expression of alx-3 at anterior facing wounds, as well as ectopic alx-3 expression near posterior regions of tail fragments associated with ectopic brain tissue. In contrast, knockdown of APC or notum, which results in higher Wnt signaling, significantly reduced wound-induced alx-3 expression at anterior-facing wounds by whole-mount in situ hybridization (Fig. 7C) and qRT-PCR (Fig. S10).

Owing to the feedback between Wnt signaling and alx-3, we reasoned that a low-Wnt environment in β-catenin(RNAi) may be able to rescue the headless phenotype of alx-3(RNAi) tail fragments. To test this, we performed double RNAi using β-catenin(RNAi) with alx-3(RNAi). Knocking down β-catenin in tail fragments resulted in anterior brain regeneration similar to controls visualized by fluorescence in situ hybridization with probes against pc2 and gpas. Double knockdown of β-catenin with alx-3 resulted in a significantly higher proportion of tail fragments that were able to regenerate a brain [93% compared with 53% in alx-3(RNAi) animals; Fig. 7D]. Interestingly, knockdown of β-catenin is already known to ablate wound-induced notum expression, and knocking down both β-catenin and alx-3 did not recue notum expression 24 HPA (Petersen and Reddien, 2011) (Fig. 7E). However, knockdown of APC increases wound-induced alx-3 expression, and knocking down both alx-3 and APC rescued this phenotype, Therefore, rescue of the alx-3(RNAi) phenotype when β-catenin expression is depleted occurs through a mechanism that does not involve rescuing wound-induced notum expression. Taken together, the rescue of the alx-3(RNAi) phenotype by inhibition of Wnt signaling suggests that alx-3 is in a negative-feedback loop regulated by Wnt signaling.

In this study, we functionally tested the aristaless-like homeobox-3 (alx-3) transcription factor that was originally found to be expressed in a subpopulation of dividing piwi-1⁺ cells called ν-neoblasts (Molinaro and Pearson, 2016). We found that alx-3 was required to maintain gad⁺ and th⁺ neurons as well as maintain brain size in proportion to body size in uninjured animals. We expected more specific effects on regenerating neurons, but were surprised to find that alx-3 was induced in stem cells located at anterior-facing wounds during regeneration, and that tail fragments could no longer regenerate heads. Defects in head regeneration could be due to failure of neurogenesis or to a defect in re-establishing anterior polarity, thereby leading to a failure in regeneration of all anterior structures. To distinguish which of these mechanisms were affected, we performed sagittal amputations to remove neural tissue while maintaining the AP axis. alx-3(RNAi) animals that maintained full AP polarity were able to regenerate missing neural tissue, suggesting the regeneration defect was not due to failure of neurogenesis specifically. We then used bulk RNAseq to uncover the pathways that were dysregulated in regenerating alx-3(RNAi) tails. We found that the alx-3 headless tail phenotype was the result of an overall failure to initiate the early wound response transcriptional programs involving both wnt1 and notum. The effect of early-acting wound response gene expression on wnt1 and notum expression is currently unclear as knockdown of several of these early-acting genes failed to produce significant regeneration phenotypes (Wenemoser et al., 2012). This would suggest that multiple early wound response genes have redundant function to activate downstream wnt1 and notum expression, and that alx-3 may be one of the early factors to activate wnt1 and notum either directly or indirectly. However, this hypothesis would be difficult to test, as we are unable to inhibit the expression of all early-acting genes involved in this wound response simultaneously. As PCGs, including wnt1 and notum, are expressed in muscle cells, we investigated whether reduced expression of wnt1 and notum in alx-3(RNAi) animals is due to dysregulation of the muscle population. However, we did not observe any significant difference in the proportion of collagen⁺ muscle cells nor any perturbation in muscle fiber morphology in alx-3-depleted animals during regeneration or in uninjured animals. This suggests that alx-3(RNAi) inhibits wnt1 and notum expression without affecting the number or morphology of muscle cells.

It is known that notum expression at anterior-facing wounds is crucial for establishing anterior regeneration polarity. Because knockdown of either alx-3 or notum results in impaired head regeneration, this suggests that alx-3 promotes notum expression in muscle cells. We put forth a model (Fig. 7F) that alx-3 activates notum expression, either directly or indirectly, in muscle cells in a cell-autonomous manner. Expression of notum is then transmitted to neighboring stem cells at anterior-facing wound sites to subsequently activate alx-3 expression cell non-autonomously to facilitate anterior regeneration.

Neoblasts are a pluripotent class of stem cells as they are constitutively active and give rise to all missing tissues during regeneration. Previous work has shown that at least some stem cells express Wnt signaling receptors, suggesting that they are capable of receiving Wnt signaling at homeostasis (Currie et al., 2016). The expression of alx-3 within stem cells at anterior-facing wounds serves as a distinguishing factor between stem cells located at anterior-facing versus posterior-facing wounds. Before this work, notum expression was considered the primary differentiating factor between anterior-facing and posterior-facing wounds, although much of its expression is restricted to collagen⁺ muscle cells (Scimone et al., 2017; Witchley et al., 2013). The regulation of alx-3 expression in stem cells by known PCGs serves as an important relationship between positional signaling that is generally restricted to muscle cells and stem cells to guide cell fate and the regenerative outcome.

alx-3 is an aristaless-like paired homeobox transcription factor (Qian et al., 1992). Mice and humans have three Alx genes: Alx1, Alx3 and Alx4; Alx3 has been lost independently in some vertebrate lineages, including frogs (X. tropicalis), lizards (A. carolinensis) and chickens (G. gallus) (McGonnell et al., 2011). Although Alx genes derived their names from the Drosophila melanogaster homolog aristaless, Alx-like genes are present across the metazoans (McGonnell et al., 2011; Ryan et al., 2006). The mouse Alx3 homolog is expressed in the cephalic mesenchyme and lateral mesoderm, and is developmentally important for neural tube closure and craniofacial development (Beverdam et al., 2001; Lakhwani et al., 2010). In zebrafish, Alx3 has a well-established role in craniofacial development: Alx3 is enriched in frontonasal neural crest cells, and loss of Alx3 results in severe neurocranium development defects (Mitchell et al., 2021; Yoon et al., 2022). In humans, homozygous mutation of Alx3 has been associated with a form of frontonasal dysplasia known as frontorhiny (Twigg et al., 2009). Across several organisms, and including this study, alx-3 plays an important role in establishing tissue identity and in the development of anterior structures, perhaps indicating that the role of alx-3 in anterior tissue patterning in planarians may be a conserved function of this gene family. It will be interesting to determine whether Alx genes are upstream of notum and Wnt signaling in other systems during embryonic development.

Asexual populations of Schmidtea mediterranea (strain CIW4) were maintained as previously described (Alvarado et al., 2002). Planarians were kept at 18°C in milliQ water supplemented with 0.21 g/l Instant Ocean, 0.1 mM KCl, 0.1 mM MgSO₄ and 0.12 mM NaHCO₃. Animals were fed calf liver paste approximately once per week.

Planarian transcripts were cloned using forward and reverse primers to generate double-stranded RNA (dsRNA) from a pT4P expression plasmid. Cloning of alx-3 was performed using an expression vector containing the alx-3 gene (SmedASXL_006490, dd_Smed_v6_11150_0_1) (Rozanski et al., 2019) provided by Molinaro and Pearson (2016). All genes were cloned using forward and reverse primers into T4P vectors as previously described (Rink et al., 2009). Primers were generated using the freely available web tool Primer3 Web version 4.1.0 (https://primer3.ut.ee). To generate RNAi food, HT115 bacterial cultures expressing dsRNA were prepared using clones expressing the pT4P vector mixed with calf liver. RNAi experiments were performed as originally described (Newmark et al., 2003), with updated modifications (Adler and Alvarado, 2018). Briefly, bacteria were grown to an OD₆₀₀ of 0.8 and induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 2 h at 37°C with shaking. Bacteria were pelleted and mixed with calf liver paste at a ratio of 500 μl of liver per 100 ml of original culture volume. Bacterial pellets were thoroughly mixed into the liver paste and frozen as aliquots at −80°C. GFP was used for all negative controls as previously described (Cowles et al., 2013). In all homeostatic experiments and regeneration experiments, RNAi food was fed to 7-day starved worms every third day for a total of nine feedings. In the homeostatic experiments, animals were fixed 10 days after the 9th feed. Amputations were performed 4 days after the 9th feed unless noted otherwise. All animals used for staining were 3-6 mm in length in the case of wild-type and homeostatic experiments, and 1-4 mm in the case of regenerating fragments, and were size matched between experimental and control worms.

Primer set AA18 (5′-CATTACCATCCCGCCACCGGTTCCATGG-3′) and PR244F (5′-GGCCCCAAGGGGTTATGTGG-3′) were used to generate a PCR template containing one T7 promoter site in the antisense direction of the gene of interest inserted in a pT4P vector as described previously (Currie et al., 2016). Antisense riboprobes were made using T7 RNA-polymerase (ThermoFisher Scientific, EP0111) over a 2-4 h incubation period. However, in the case of weakly expressed genes such as alx-3, an overnight incubation of the DIG probes resulted in optimal detection of the gene transcript.

Riboprobes were made from PCR templates using the pT4P vector generated above. Whole-mount and fluorescent in situ hybridization experiments were performed as previously described (Currie et al., 2016; Pearson et al., 2009) with the following modifications that optimized results. Briefly, 5% N-acetylcysteine in phosphate-buffered saline (PBS) was used to kill the worms and remove mucus, followed by fixation in 4% formaldehyde in PBST (0.3% Triton-X) for 20 min. Worms were rinsed with PBST and further permeabilized with reduction solution consisting of 50 mM DTT, 1% NP-40 and 0.5% SDS in PBS for 3-5 min at room temperature. Worms were dehydrated with methanol and stored at −20°C then bleached with 6% hydrogen peroxide (in methanol) overnight and rehydrated with PBST. Worms were pre-hybridized for 2 h at 56°C then hybridized with probe overnight at 56°C. Blocking solution (10% horse serum in MABT) was used for blocking and antibody incubation. Colorimetric stains (whole-mount in situ hybridization) were developed using 4-nitro blue tetrazolium chloride (NBT, Roche, 11383213001) and 5-bromo-4-chloro-3-indolyl-phosphate (BCIP, Roche, 11383221001). Fluorescence in situ hybridization stains were developed with either tyramide amplification in borate buffer, or Fast Blue B salt (Sigma D9805) with naphthol AS-MX phosphate (Sigma 855). For immunostaining, rabbit anti-H3ser10p (H3P) was used at 1:1000 (EMD Millipore, 05-817R-I) or mouse anti-PIWI-1 (kind gift from Dr Jochen Rink, Max Planck Institute for Multidisciplinary Sciences, Germany) was used at 1:1000 (Zhu and Pearson, 2018). BrdU (Sigma-Aldrich B5002) labeling was performed as previously described (Zhu et al., 2015). Briefly, worms were adapted to a high salt medium (5 g/L Instant Ocean) for 1-3 days before BrdU delivery and maintained in high salt for the duration of the chase period. BrdU dissolved in 50% DMSO was fed at a concentration of 10 mg/ml in liver paste. Fixation and fluorescence in situ hybridization were performed as described above. After fluorescence in situ hybridization, worms were incubated in acid for 45 min (2 N HCl and 0.5% Triton-X), then neutralized with 0.1 M sodium borate. Blocking solution (10% BSA and 5 mM thymidine in PBST) was used for blocking and antibody incubation. BrdU was detected with mouse anti-BrdU at 1:300, followed by anti-mouse HRP (1:500) and tyramide amplification. Only animals that exhibited staining throughout the entire body were quantified.

Colorimetric whole-mount in situ hybridization stains were imaged on a Leica M165 fluorescent dissecting microscope. Fluorescent stains were imaged on a spinning disk confocal microscope (Olympus IX81S1F-3) with a Hamamatsu C9100-13 EM-CCD camera and a Yokogawa CSU X1 scan head, employing Perkin Elmer Volocity software. All image quantifications and post-processing were made using the freely available ImageJ software Fiji (http://rsb.info.nih.gov/ij/) and Imaris (http://www.bitplane.com/imaris/imaris). All experiments were, at minimum, triplicated and at least 10 worms were used per stain and per time point (i.e. n≥30). All statistical analyses between RNAi groups were carried out using a two-tailed Student's t-test. All images were post-processed using Adobe Photoshop and figures assembled in Inkscape.

Reverse transcription reactions were conducted on total RNA extracted from ∼10 whole worms or 20 tail fragments using Trizol Reagent (Thermo Fisher) a SuperScript III Reverse Transcriptase Kit (Invitrogen). Quantitative real-time PCR was performed in biological triplicate on a Bio-Rad CFX96 Touch Real-Time PCR Detection System with SYBR Green PCR Master Mix (Roche) as per the manufacturer's instructions. Expression was normalized to control(RNAi) and the 2^−ΔΔCT method was used for relative quantification. Primer pairs for ubiquitously expressed GAPDH were used as a reference as previously described (Eisenhoffer et al., 2008). Experiments were carried out in biological and technical triplicates. All statistical analysis was carried out by performing Welch's unequal variances t-test using Microsoft Excel. *P<0.05, **P<0.01 and ***P<0.001. Graphs were generated using GraphPad Prism. All error bars indicate the s.d.

RNA deep sequencing (RNAseq) was performed on whole and amputated tail fragments 4 days after nine RNAi feedings for control(RNAi) and alx-3(RNAi). RNA was extracted using Trizol RNA Extraction Kit (Thermo Fisher). Experiments were performed in biological triplicates and sequenced to a depth of ∼30 million reads per sample and multiplexed on an Illumina NovaSeq with 50 bp paired-end reads. Primer sequences were removed using trimmomatic and sequences were then aligned to a previously described planarian transcriptome Smed_ASXL in NCBI Bio Project PRJNA215411 using the program Salmon (https://bioconductor.org/packages/devel/workflows/vignettes/rnaseqDTU/inst/doc/rnaseqDTU.html; Love et al., 2018). DESeq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html) (Love et al., 2014) was used to determine significantly up- or downregulated transcripts in control(RNAi) and alx-3(RNAi) animals. Heatmaps were generated using the pheatmap R package in RStudio. DESeq2 output on processed data for all time points are in Table S1.

We thank Dr Alyssa M. Molinaro for help making the single cell plots in Fig. 1. We thank the Imaging Facility at The Hospital for Sick Children.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201777.reviewer-comments.pdf.