Amputation of a salamander limb triggers a regeneration process that is perfect. A limited number of genes have been studied in this context and even fewer have been analyzed functionally. In this work, we use the BMP signaling inhibitor LDN193189 on Ambystoma mexicanum to explore the role of BMPs in regeneration. We find that BMP signaling is required for proper expression of various patterning genes and that its inhibition causes major defects in the regenerated limbs. Fgf8 is downregulated when BMP signaling is blocked, but ectopic injection of either human or axolotl protein did not rescue the defects. By administering LDN193189 treatments at different time points during regeneration, we show clearly that limb regeneration progresses in a proximal to distal fashion. This demonstrates that BMPs play a major role in patterning of regenerated limbs and that regeneration is a progressive process like development.
The ability of salamanders to regenerate their tissues (limbs, skin, spinal cord, etc.) has been studied for over 250 years, with experiments ranging from physiological observations (Spallanzani, 1769) to transplantations (Maden, 1980), to the more recent high-throughput transcriptome and genome sequencing (Stewart et al., 2013; Nowoshilow et al., 2018). Despite so much work on describing the phenomenon of tissue regeneration, very few of its inner workings are known. Among them, BMP signaling has been identified as a crucial pathway in regeneration of mice digit tips (Han et al., 2003; Yu et al., 2010), Xenopus limb buds/tail (Beck et al., 2006) and zebrafish fins (Smith et al., 2006). This pathway has also been found to be important in axolotl (Ambystoma mexicanum) limb regeneration (Guimond et al., 2010), but the understanding of its role in this context remains limited.
During normal development, BMP signaling has known roles in patterning, control of cell survival and proliferation, as well as osteogenesis (Luo et al., 1995; Furuta et al., 1997; Macias et al., 1997; Zuzarte-Luis et al., 2004; Hamaratoglu et al., 2014; Salazar et al., 2016). Bmp2, Bmp4, Bmp5 and Bmp7 are all expressed in the developing limb (Geetha-Loganathan et al., 2006), and, except for Bmp5, have also been studied in mice digit tip regeneration. In neonatal mice, digit tip regeneration is limited to the region expressing Msx1. Msx1−/− mice do not regenerate, but that phenotype is rescued by implanting BMP4-soaked beads (Han et al., 2003). In addition, BMP7-soaked beads extend the region normally capable of regeneration beyond the first phalange in normal mice, a process that involves re-expression of endochondral markers, instead of direct ossification (Yu et al., 2010). In axolotls, BMP2 is the only one that has thus far been studied in regenerating limbs and has been shown to affect proliferation and apoptosis during the process (Guimond et al., 2010), similar to BMP4 inducing proliferation in mice digit tip regeneration (Han et al., 2003). These facts indicate a crucial role for these genes in regeneration. Global manipulation of BMP signaling should therefore have a profound impact on the amphibian limb regeneration process.
Urodele limb regeneration leads to the complete replacement of a normal functional structure following amputation. It would thus be expected that some parts of the process would be similar, at least to some degree, to those observed during development. However, in contrast to development, limb regeneration initiates from mature tissue and can start from any proximo-distal position along the limb (Goss, 1969; Iten and Bryant, 1973). No matter where the amputation is performed, regenerated tissue converges towards the same end result: replacing exactly what is missing. This has led to one of the main questions regarding regeneration: how do regenerating limbs manage to regrow exactly the missing structure? Historically, researchers have presented two models to explain the accuracy of regeneration: cell intercalation and progressive specification (French et al., 1976; Bryant et al., 1981; Echeverri and Tanaka, 2005; Roensch et al., 2013). Although evidence supporting each model has been published, recent works point towards the progressive model as the actual process governing normal regeneration (Roensch et al., 2013).
The intercalation model claims that regeneration is induced by the opposition of positional identities that are not normally in contact. Their contact would stimulate the filling of the ‘identity gap’ between them. This is best demonstrated in cockroach, where an inverted leg segment can be regenerated between a proximally amputated leg grafted to a distal leg stump (French et al., 1976). In planarians, intercalation can be induced with grafts that cause opposition on either the dorso-ventral or antero-posterior axis, resulting in supernumerary structures (i.e. additional pharynxes) (Agata et al., 2003). In normal salamander regeneration, the opposing structures would be a proximal stump and a wound epithelium or apical epidermal cap that adopts a distal identity. New growth from both proximal and distal tissue would simultaneously be specified, followed by the intermediate structures. Indeed, grafting of a distal blastema to a proximal stump allows the regeneration of all intermediate proximo-distal elements, no matter what the original stage of the blastema (Iten and Bryant, 1975). It has also been observed that HoxA13 mRNA, an autopod associated gene, is detected as early as 24 h post amputation, regardless of whether the amputation is through the autopod or stylopod (Gardiner et al., 1995).
The progressive model posits that regeneration proceeds exclusively in a proximal to distal fashion, as in developing limbs (Saunders, 1948). The wound epithelium is still necessary (Thornton, 1957) but would not be responsible for positional identity of cells within the blastema. This model is supported by a recent study by Tanaka's group, who observed that the expression of HoxA9-13 proteins in regenerating limbs has a timing similar to that of developing limbs (Roensch et al., 2013), contrary to earlier work (Gardiner et al., 1995). Supporting the progressive model, recent work suggests that the base of the blastema acquires a stable identity while the apex is still labile (McCusker and Gardiner, 2013).
Despite all the research on these models, the mechanisms through which signaling pathways contribute to the process of progressive vs intercalation regeneration are still elusive. In this work, we used the chemical inhibitor LDN193189 to explore the role of BMP signaling in regeneration. LDN193189 inhibits the phosphorylation of ALK2, which is required for canonical signal transduction from the extracellular BMPs to the intracellular Smad1 and Smad5 (Vogt et al., 2011). Although this drug does not target specific members of the BMP family, it has a widespread effect on all BMP signaling. In the animals, it is used with consistency and temporal precision to specifically block BMP signaling. With this tool, we show that BMP signaling is a key player in patterning and there is strong evidence that cell patterning of the limb segments in regeneration is a progressive proximo-distal process.
Gene expression during normal limb regeneration
The stages of regeneration are represented in Fig. 1A, along with their abbreviations and the corresponding time required to reach them (indicated as days post-amputation or DPA). Named regeneration stages are based on Tank's staging system (Tank et al., 1976). The time required to reach those stages can vary with the size of the animals and environmental factors (Schauble, 1972; Schauble and Nentwig, 1974; Wallace, 1981), and were determined experimentally in our animals. The time required to reach specific stages has been consistent within our cohorts.
The expression levels of various BMPs and some of their known targets were measured during regeneration using qRT-PCR (Fig. 1B). Although Bmp4 has been shown to be important in mice digit tip regeneration (Han et al., 2003), it has never been cloned in axolotl and is absent from the genome assembly (Nowoshilow et al., 2018), so we can only look at its closest paralog, Bmp2 (Feng et al., 1994). Bmp2, Bmp5 and Bmp7 all show notable rises in expression starting at the early bud (EB) stage. Their expression remains high all the way through the early differentiation (ED) stage, covering the entire redevelopment phase of regeneration. Interestingly, these BMPs do not increase in the preparation phase of regeneration, which we have previously shown to correspond to a TGFβ1 increase (Lévesque et al., 2007).
Potential BMP target genes were chosen for their roles in different processes of development or regeneration. Msx1 and Msx2 are involved in proliferation and maintenance of non-differentiated state of cells (Bendall et al., 1999; Bendall and Abate-Shen, 2000; Odelberg et al., 2000; Hu et al., 2001). Shh controls antero-posterior patterning (Riddle et al., 1993; Roy and Gardiner, 2002). Fgf8 and Fgf10 participate in maintaining limb bud outgrowth during development (Crossley et al., 1996; Ohuchi et al., 1997; Xu et al., 1998). Sox9 and collagen 2A1 (Col2a1) are involved in chondrogenesis and osteogenesis (Bi et al., 1999; Akiyama et al., 2002; Pan et al., 2008). These target genes see an increase in expression at EB, similar to BMPs, although the specific pattern varies depending on the genes (Fig. 1B). Shh, Fgf8 and Fgf10 expressions peak around medium bud stage (MB) then decrease, while Msx1, Msx2, Sox9 and Col2a1 peak at the later stage of palette stage (Pal). These results are consistent with the fact that specific blastema stages are associated with distinct transcription patterns (Knapp et al., 2013; Voss et al., 2015). In addition, later stages of regeneration correspond to a progression in the morphogenetic process and some of these genes are directly associated with the establishment of tissues, such as skeletal elements, muscle and skin.
The expression of Bmp2, Bmp7, Fgf8 and Fgf10 were localized by in situ hybridization (Fig. 1C-F). They are all expressed throughout the mesenchyme of the blastema but Fgf8 is slightly more intense in the region closer to the AEC (apical ectodermal cap, the epithelium covering the blastema). Only Bmp7 is detected in the AEC itself.
BMP signaling and inhibition by LDN193189
BMP canonical signaling was assessed by measuring Smad1/5 phosphorylation during the various stages of limb regeneration. Total Smad1 was detected starting at 48 h, while the increase of total Smad5 was mostly seen from EB onwards (Fig. 2A). Our result show that phosphorylation of Smad1/5 is strong and stable from EB until ED (Fig. 2A,A′). This corresponds exactly to the stages of regeneration where BMPs expression is increased. At both EB and late bud stage (LB), phospho-Smad1/5 was localized in the nuclei of the AEC, of mesenchymal blastema cells, as well as in the nuclei of the transected end of muscles (Fig. 2C,C′,G).
LDN193189, a BMP signaling inhibitor, was selected to test whether it was possible to block phosphorylation of Smad1/5 in axolotls. The inhibition of Smad1/5 phosphorylation by LDN193189 has already been shown in vitro (Vogt et al., 2011) and in vivo (Cannon et al., 2010), but not specifically for axolotl regeneration. We therefore selected a stage where Smad1/5 is strongly phosphorylated (MB) and treated animals with different doses for 48 h before that stage (Fig. 2B,B′). Our data clearly showed a dose-dependent inhibition of Smad1/5 phosphorylation by LDN193189. The reduction of phospho-Smad1/5 detection was also seen by immunofluorescence for all tissues where nuclei are positive in controls (Fig. 2C-J) The range of 0.2 µM to 1.0 µM LDN193189 was judged to be appropriate to study different levels of BMP signaling inhibition, as these concentrations inhibit Smad1/5 phosphorylation without affecting the health of animals or their ability to feed and grow. The specificity of LDN193189 for canonical BMP signaling in axolotls was tested by looking into the phosphorylation of ERK1/2 and Smad2. Neither was affected by the concentrations of LDN193189 used in our experiments (Fig. S1A-B′). The lack of effect on the Smad2 pathway was further supported by the appearance of a blastema in all LDN193189-treated animals, whereas no blastema is formed when Smad2 phosphorylation is inhibited (Denis et al., 2016).
Effect of LDN193189 treatments on limb regeneration
To assess the impact of BMP signaling inhibition, we treated animals with LDN193189 throughout the entire process of regeneration (Fig. 3). The first noticeable effect, around EB-MB (Fig. 3B,G,L,Q), was that blastemas are shorter than in controls. This tendency was maintained for the rest of the regeneration process. In treated animals, the elongation of the blastema eventually slowed down and stopped around Pal (Fig. 3H,M). At 0.2 µM and 0.5 µM, there was still some phosphorylation of Smad1 and Smad5, which explains the partial growth of the blastema past EB. In the following weeks, there was a regression of the blastema in the animals treated with 1 µM LDN193189 (Fig. 3Q-S). Once regeneration was completed in controls, the animals were fixed and stained for the formation of skeletal elements (still cartilage at that age) (Fig. 3E,J,O,T). There were clearly a number of skeletal elements missing in animals treated with LDN193189, an effect that was dose dependent, with higher concentration leading to less cartilage formation. At 1 µM, no new skeletal element formation was seen and the radius and ulna often ended up shorter than they were at the time of amputation. As a reference, the normal skeletal structure of an axolotl forelimb is displayed in Fig. 3U.
Impact of BMP signaling inhibition on cell proliferation and survival
To assess whether the missing structures were simply due to a low cell number, we looked into proliferation and apoptosis. Animals were treated with LDN193189 starting 24 h before amputation and limbs were harvested for histology at the EB and LB stages (Fig. 4). We assessed proliferation using BrdU incorporation and apoptosis using the TUNEL assay. The percentages of nuclei positive for BrdU and TUNEL are compiled in Fig. 4I-L. At EB, there was abundant proliferation in all animals (26-33% of cells, no significant difference between treatments), localized mostly close to the wound epithelium. At this stage, an increase in apoptosis could be seen in LDN193189-treated animals but only at higher concentrations (up to 1.5% of cells). The minimal effect of inhibition of BMP signaling on cell proliferation was consistent with the observation that the role of BMPs starts from EB onwards. At LB stage, a decrease from 43% to 14% in the number of proliferating cells coincided with a smaller blastema in a concentration-dependent fashion. Apoptosis followed the opposite trend with an increase in TUNEL positive cells as the dose of LDN193189 increases (2.3% at 1 µM from 0.65% in DMSO controls). This indicates that blocking BMP signaling lowers both proliferation and cell survival.
Impact of BMP signaling inhibition on gene expression
We selected the stages typically associated with redevelopment (EB to Pal) and tried to determine the impact of BMP signaling inhibition on gene expression (Fig. 5). Animals were treated with LDN193189 starting 48 h before the stage when the RNA was harvested (see Fig. 1A for timing used) and expression was measured by qRT-PCR. The changes in expression depended on the dose of drug and on the stage when the drug was administered. The first noticeable result was that the expression levels of the BMPs themselves could change when their signaling is blocked. Bmp2 was mostly affected at later stages, where a rise in expression was observed (LB and Pal), whereas Bmp5 was mostly affected earlier with a decrease in expression at EB. The effect of LDN193189 on Bmp7 was weak and inconsistent across doses and stages of regeneration.
As expected, many of the BMP target genes have their expression reduced in a dose-dependent manner by LDN193189 treatment. However, the scale of reduction depends on the specific stage. In the case of Msx1, the expression reduction is very pronounced until the Pal stage, at which point the expression becomes almost unaffected. The same is true for Msx2 and Fgf8, but to a lesser extent. Shh expression is also inhibited, but mostly at EB. Sox9 expression is reduced by LDN193189 throughout regeneration whereas Col2a1 is mostly affected at Pal. The treatments cause only a weak change in the expression of Fgf10. These results confirm that most of these genes are downstream of BMP signaling at specific points during limb regeneration, as demonstrated in other systems where they are expressed in a similar way (Hogan, 1996).
Rescue of LDN193189-treated limbs by injection of FGF8 protein
LDN193189 treatments cause a reduction in Fgf8 expression and, at 0.2 µM, causes phenotypes similar to those observed in Fgf4;Fgf8 double KO, where the radius and anterior digits are partially or completely missing (Sun et al., 2002). This points towards Fgf8 being important for the role of BMP in patterning or at the very least in maintaining growth. We therefore attempted to rescue the phenotype resulting from BMP inhibition by injecting recombinant human or axolotl FGF8 protein into blastemas during the period when it is normally expressed. In this experiment, 0.5 µM LND193189 treatment was administered from DPA5 to DPA12 (from 2 days before EB, until LB), which correspond to the times when FGF8 expression peaks in normal regeneration. Treatments were stopped (to allow osteogenesis) at LB in order to observe whether limbs injected with FGF8 would regenerate skeletal elements. In LDN193189-treated limbs injected with PBS or FGF8, no consistent improvement in regeneration was observed in either types of injections (n≥20).
Impact of delaying LDN193189 treatment on regenerating limbs
BMPs have known roles in modulating proliferation, patterning and bone induction (Hogan, 1996). As seen with our qRT-PCR results, the downstream genes associated with these processes are expressed at different times during regeneration. We therefore assumed these processes would act in a succession and attempted to inhibit them individually by initiating LDN193189 treatment at various stages. Skeletal preparation of regenerating limbs treated with LDN193189 starting at different times post-amputation are shown in Fig. 6. As seen in Fig. 3, a higher concentration of drug results in more severe phenotypes. Delaying initiation of treatments shows another clear tendency: the later the treatment starts, the more skeletal elements are regenerated. Identical results are also observed when limbs are amputated through the middle of the stylopod instead of the zeugopod (Fig. S3). More importantly, the skeletal elements are always added in order from proximal to distal (especially clear for 0.5 and 1 µM doses), demonstrating that limb regeneration is progressive.
There is also an effect on the antero-posterior axis, which can be seen when comparing 0.2 µM treatments started at EB, MB and LB (Fig. 6G-I). Although those amputated limbs all regenerated up to the first phalanx, the anterior-posterior patterning of the limb was partially re-established when the treatment was started later. In treatments starting at EB, only one out of eight limbs regenerated both radius and ulna (five regenerated only one of them and two showed no regeneration). In those started at MB, six out of eight limbs regenerated both radius and ulna, as well as some carpal elements. Starting treatment at LB will again allow extra skeletal elements to form. Four out of eight limbs had all carpal cartilages and some digit cartilages. The other four had half the number of carpals and one or two digit cartilages. Starting treatment at Pal allowed regeneration of all eight carpal elements in eight out of eight limbs. In six out of those eight limbs, a complete second finger was regenerated (identified by its position relative to the carpals) and part of a third finger.
As delaying LDN193189 treatment allows the regeneration of an increasing number of proximal to distal skeletal elements, we investigated whether the expression of Hoxa11 and Hoxa13 is affected in these animals (Fig. 7). The DMSO controls had a stable expression of Hoxa11 throughout regeneration, except at LB, where it is 40% higher than other stages. A LB peak was also observed with Hoxa13, this time with an expression 8.5× higher than at EB. At EB, both genes were expressed in the mesenchymal cells close to the amputation site. They were still expressed in the mesenchymal cells at LB but, compared with Hoxa11, Hoxa13 was slightly restricted to the distal part of the blastema. LDN193189 treatments caused a reduction of expression of both genes, depending on the specific time point measured. Hoxa11 expression was up to 38% lower at EB, but only 20% lower at MB and 15% at LB and Pal. Similarly, Hoxa13 was reduced by as much as 60% at EB, 38% at MB, 34% at LB and 18% at Pal. The treatments reduced the expression level of these genes but had little effect on their localization. The most noticeable was Hoxa13 at LB, where expression is mostly lost in the proximal part of the blastema.
BMP signaling regulates the expression of multiple genes during regeneration
Regeneration leads to a limb identical to the one produced during limb development. It is therefore expected that the two share some of the same processes. Our results show that multiple development-related genes are expressed at specific times during regeneration. These expression patterns are consistent with the idea that, like development, regeneration involves sequential patterning and differentiation events. BMPs are responsible for the induction or maintenance of many genes at multiple time points and are therefore necessary for proper limb formation.
The dependence of most of these genes on BMP signaling has been demonstrated previously in developmental studies. Here, we demonstrate for the first time that it is similar in the context of regeneration. Interestingly, the dependency on BMP signaling itself varies on the stage observed, a fact most notable with Msx1, the expression of which is inhibited with LDN193189 only before it reaches its peak at Pal. Here, we see BMPs as being upstream of Msx1, but other studies have shown BMPs to be downstream of Msx1, like Bmp4 is in mice digit tip regeneration (Han et al., 2003). This could possibly be a species-specific difference or a difference between the process of limb regeneration in axolotls and digit tip regeneration in mammals.
The relationship between BMP signaling and Fgf8 expression is complex. Early in limb bud development, BMPs are required for induction of FGFs in the AER but downregulate them later (Fernandez-Teran and Ros, 2008; Benazet et al., 2009; Zeller et al., 2009). During regeneration, our results also show a dependency of Fgf8 on BMP signaling. The effect is strongest at the EB stage, weakens at MB and LB and is not significant at Pal. We see no indication that inhibition of BMPs ever increases Fgf8 expression, as would be suggested by the Shh-Grem1-Fgf feedback loop model established for developing limbs (Zeller et al., 2009). As the expression of Fgf10 changes very little in response to LDN193189 treatments, we also lack evidence of the Fgf8 and Fgf10 mutual induction that is observed in chicken limbs (Ohuchi et al., 1997). Our results regarding Fgf10 and Fgf8 expression are basically identical to those obtained by Christensen et al. (Christensen et al., 2001, 2002), which do not indicate that these genes cooperate in urodele limb regeneration as they do in amniote development. We did not present Fgf4 because we see, as Christensen et al. did, a decrease following amputation (data not shown), and we therefore concluded that it is not a major player during the limb regenerative process in axolotls. Again, these differences could be species specific.
BMPs are also required for a precise expression domain of Fgf8 in developing limbs. In mice, AER-specific Bmp2/Bmp4/Bmp7 triple knockouts causes an expansion of the Fgf8 expression domain followed by an early arrest of expression, leading to a polydactyly with shortened fingers (Choi et al., 2012). Expression of Msx1 and Msx2 is also much weaker in those knockouts, fitting with our own inhibition results. In developing chicken limbs, BMP inhibition with Noggin makes the expression region of Fgf8 narrower and irregular. Ectopic expression of Msx1 in the dorsal ectoderm induces misexpression of Fgf8 (Pizette et al., 2001), suggesting that BMPs promote Fgf8 expression through Msx1. This could indicate that the reduction in Fgf8 expression seen in LDN193189-treated animals is indirect. In turn, the reduced Shh expression could also be from the reduced induction by Fgf8.
Among the genes we looked at, Sox9 is the one most directly involved in chondrogenesis. Its expression rises, starting at EB, and is susceptible to dose-dependent inhibition by LDN193189 treatment at all time points tested. Sox9 inhibition is the likely reason for the missing skeletal elements resulting in truncated limbs. In developing mice, limb-specific deletions of Sox9 completely prevent cartilage formation, while Col2a1-Cre (chondrocyte-specific) deletions allow it, although they are much shorter than in wild-type mice. This is explained by the sequential roles of Sox9 in both mesenchymal cell differentiation into chondrocytes and chondrocyte proliferation (Akiyama et al., 2002). Low dose LDN193189 treatment allows cartilage regeneration, indicating functional chondrocyte differentiation and proliferation, which is explained in part by the presence of some phospho-Smad1/5. At higher doses, there is still some expression of Sox9, even though regeneration is inhibited. This could be the result of having reached a threshold of Sox9 expression that is not sufficient to sustain chondrogenesis. It could also be due to the other genes affected by the marked reduction in the levels of phospho-Smad1/5.
BMP signaling regulates AP and PD skeletal patterning in the limb
In developing limb, expression patterns of BMPs depend on the stage of development, in addition to the specific BMP itself. In chicken and mice, they tend to first be expressed in the AER, as well as the posterior or anterior mesenchyme. This expression region then expands to cover the entire region adjacent to the AER (Geetha-Loganathan et al., 2006; Choi et al., 2012). As the fingers later become specified, BMPs are restricted to the contour of presumptive skeletal elements, each with their specific pattern (Lorda-Diez et al., 2014), implying a more-precise patterning. In axolotl development, Bmp2 expression was detected towards the end of limb development (Guimond et al., 2010). Contrary to chicken and mice, its expression is anterior instead of posterior, corresponding to the order of formation of the skeletal elements, which appear from anterior to posterior in axolotls (Bandyopadhyay et al., 2006; Geetha-Loganathan et al., 2006; Guimond et al., 2010). In contrast to development, Bmp2 expression during regeneration first covers the entire blastema. It is only from LB onwards that its expression becomes more restricted, mirroring the development pattern (Guimond et al., 2010). We observe that Bmp7 is also expressed in the entire blastema during the EB and LB stages. Even with an expression that is not restricted to small regions of the mesenchyme, blocking BMPs signaling during these stages leads to specific phenotypes, especially at 0.2 µM LDN193189.
The requirement of BMP signaling translates into complex phenotypes when regenerating animals are treated with LDN193189. The reduced cell proliferation and survival can certainly play a role, but as an abnormal limb is produced instead of a miniature hand, defective patterning is a more likely problem. BMP signaling first affects patterning, then the osteogenesis process itself. A low dose of LDN193189 inhibits normal patterning, while allowing some of the cartilages to form (Fig. 6F-J). This indicates that these processes have different levels of sensitivity to BMP signaling, as high concentrations of LDN193189 block all cartilage formation.
The redevelopment phase of regeneration was generally considered to start at LB, an assumption based on the nerve requirement for regeneration (Grim and Carlson, 1979; Wallace, 1981). However, the antero-posterior patterning gene Shh is expressed as soon as EB and is susceptible to the inhibition of BMP signaling (Figs 1 and 5). Simply delaying LDN193189 treatment from EB to MB allows the antero-posterior axis of the zeugopod to be defined (Fig. 6G,H). This supports the idea that Shh is affected by BMP signaling, meaning the latter influences AP patterning. It also suggests that redevelopment has likely already started at EB.
Proximo-distal patterning seems to be determined in a progressive way. This is made clear by the fact that when BMP inhibition is delayed, skeletal elements are added in a proximal to distal order: radius/ulna, carpals, metacarpals and finally phalanges. Within a given antero-posterior region, no distal element is regenerated without the supporting proximal elements, which goes against the intercalation model. Each regenerated section also appears to influence the state of the next section. The impact is seen both on the proximo-distal axis and the antero-posterior axis. High-dose treatments starting at LB stage do not allow digit formation, indicating that they are not yet defined (Fig. 6). Our results show at most a 60% inhibition of Hoxa11 and Hoxa13 expression. In Hoxa13−/−; Hoxd13+/− mice, the autopod develops four partial digits and even Hoxa13−/−; Hoxd13−/− mice have some cartilaginous elements in the autopod (Fromental-Ramain et al., 1996). If even homozygous knockouts of Hoxa13 are insufficient to eliminate all cartilaginous growth at the autopod level, it is unlikely that the lower level of expression of this gene plays a crucial role in causing the phenotypes seen here.
There is strong support for Fgf8 requirement of BMP signaling in the growth of both the antero-posterior and proximo-distal axis. The removal of the AER in chicken induces cell death in the mesenchyme, and FGF8-soaked beads will lengthen fingers and even add phalanges (Sanz-Ezquerro and Tickle, 2003). Mutant mice with reduced Fgf8 expression, such as those with the Dac mutation, miss all but the most posterior finger (Kano et al., 2007). Our results provide evidence that Fgf8 signaling is also downstream of BMP signaling during limb regeneration. However, daily injections of the Fgf8 recombinant protein (either human or axolotl) are insufficient to rescue the regenerative process inhibited by LDN193189.
In this study, we show that BMP signaling is continuously active and essential during the redevelopment phase of regeneration. Its inhibition prevents normal expression of genes required for skeletal patterning on both the antero-posterior and proximo-distal axes. We also demonstrate that patterning is defined progressively, starting from proximal elements to successively more distal elements. The idea of a progressing basal stability and distal lability has been suggested before (McCusker and Gardiner, 2013), but this is the first study in which a specific signaling pathway has been manipulated to show that regeneration is indeed a progressive process.
MATERIALS AND METHODS
Animal care and surgeries
Axolotl larvae (Ambystoma mexicanum) were purchased from the Ambystoma Genetic Stock Center (Lexington, Kentucky, USA). Animals were kept individually in 20% Holtfreter's solution (0.29 mM MgSO4, 0.19 mM CaCl2, 0.17 mM KCl, 12.5 mM NaCl, 0.45 mM Trizma fish grade) at 19-22°C, with a 12 h lights-on, 12 h lights-off cycle. Prior to surgeries, animals were anesthetized by transferring them to 0.1% MS222 (Sigma, E10521, diluted in Holtfreter's solution; pH 7.0) until they no longer reacted to hind leg pinch stimulus. Both forearms were amputated from each animal (3-6 cm from snout to tip of tail) through the radius and ulna. Cartilage and tissue were then quickly trimmed to have a flat wound surface. Named regeneration stages were determined visually on the control (DMSO-treated) or untreated animals. When treatments needed to start before specific stages, a few animals underwent amputation 3 days before the experimental animals to determine the timing of stages for that cohort. When experiments were completed, animals were anesthetized and euthanized by decapitation. All care and experiments were approved by the Université de Montréal's animal care and ethics committee.
Solubilization of LDN193189 and treatments
The vials of LDN193189 (Sigma SML0559) were weighed prior and after solubilization of the drug to ensure accurate knowledge of the amount acquired. DMSO 0.5 ml per mg of LDN193189 was added to the bottle, 1 ml at a time. Each ml was mixed by pipetting up and down and transferred to another tube. The powder was then completely solubilized by pipetting up and down, and the real concentration was calculated. Stock solutions of 0.2 mM, 0.5 mM and 1 mM were then made by diluting in DMSO. Stock DMSO was used as a control. Working solutions were made by diluting stock solutions 1:1000 in Holtfreter's solution. Treated animals were transferred in working solution for the schedules specified in each experiment. LDN193189 working solutions were made fresh from stocks and changed daily.
Victoria Blue staining of cartilage
Once regeneration was completed, animals were sacrificed and fixed in Bouin's solution [71% saturated picric acid, 24% formaldehyde solution (37%) and 5% glacial acetic acid] overnight. Samples were then rinsed twice for 2 h in 70% ethanol, rinsed in 2% NH4OH until no longer yellow, bleached 30 min in 3% H2O2, soaked overnight in acid alcohol (70% ethanol and 0.37% HCl), stained 2 h in 1% Victoria Blue (in acid alcohol), thoroughly destained in 70% ethanol, dehydrated in 95% and 100% ethanol, and cleared and stored in methyl salicylate.
RNA isolation, reverse transcription and qRT-PCR
After amputation through the forearms, the animals were left to regenerate until the desired stages, with or without treatment. The animal were then anesthetized and the blastema or stump (up to 0.5-1 mm from the wound site) were harvested and homogenized in TRIzol reagent (Life Technologies, 15596018). RNA was extracted using the provided protocol, except that precipitated RNA was washed three times with 75% ethanol and was dissolved in 10 µl Ultrapure RNAse-free water (Life Technologies, 10977015). RNA was quantified by diluting 1 µl of samples in 100 µl of 2.5× SYBR Green II (Life Technologies, S7580), 5 mM Tris-HCl (pH 8.0) and reading the fluorescence (excitation: 485 nm; emission: 520 nm). The sample concentrations were determined by comparing with a pre-quantified standard.
Reverse transcription was carried out using 250 ng of RNA. RNA was mixed with 2 µl 25 µM dT17 primer, 0.5 µl 100 ng/µl random hexamers (Life Technologies, 48190011), 1 µl 10 mM dNTP (Life Technologies, 10297018) and completed to 12 µl with RNAse-free water. The mix was heated at 65°C for 5 min then put on ice for 2 min. To the mix was added 4 µl 5× First Strand Buffer, 2 µl 0.1 M DTT and 1 µl RNAseOUT (Life Technologies, 10777019), then incubated at 42°C for 2 min. Finally, 1 µl of Superscript II reverse transcriptase (Life Technologies, 18064014) was added to the mix, which was then incubated at 25°C for 10 min, 42°C for 50 min, 50°C for 10 min and 70°C for 15 min. After completion, the mix was diluted by adding 140 µl deionized water.
qPCR reactions were carried out in a White LightCycler 480 Multiwell Plate 96 (Roche, 04 729 692 001) containing 20 µl of 1× NEB Standard Taq Buffer+0.5 mM MgSO4 (NEB, B1003S), 0.33× SYBR Green I (Life Technologies, S7580), 0.2 mM dNTP, 250 nM each forward and reverse primers, 0.5 U Hot Start Taq (NEB, M0495L) and including 2 µl of sample cDNA. All primers (Life Technologies) were tested for their efficiency and are listed in Table S3. Cycling and measuring was performed using a Light Cycler 480 system with the following program: denaturation at 95°C for 2 min, quantification with 45 cycles of 94°C for 15 s, 60°C for 30 s and 72°C 60 s; melting curve at 65°C for 1 min followed by ramping to 97°C at 0.11°C/s. Analysis was carried out using LightCycler 480 software release 1.5.0 and Cp were calculated using the ‘Abs Quant/2nd Derivative Max’ analysis. Expression relative to Gapdh [appropriate as a normalization gene in salamander limb regeneration (Vascotto et al., 2005)] was calculated using Cp and amplification efficiency specific for each primer pair. The complete axolotl sequence for Shh was obtained using RACE (based on Schramm et al. (2000) and was deposited in Genbank under accession number KX809594.
Protein harvesting and western blot
Blastemas or stump (up to 0.5-1 mm from the wound site) were harvested and put into tubes containing 50 µl of modified Laemmli Buffer [97.75 mM Tris-HCl (pH 6.8), 3.75% SDS, 18.75% glycerol, 200 mM DTT and 50 mM NaF] kept on dry ice. Samples were then sonicated and quantified using the EZQ Protein Quantitation Kit (Life Technologies, R33200). Samples (20 µg in 20 µl) were run on 5% acrylamide stacking gels [125 mM Tris-HCl (pH 6.8), 0.2% SDS, 5% 29:1 acrylamide mix, 0.1% ammonium persulfate and 0.1% TEMED], 10% acrylamide running gels [375 mM Tris-HCl (pH 8.8), 0.2% SDS, 10% 29:1 acrylamide mix, 0.04% ammonium persulfate and 0.05% TEMED] and transferred overnight on PVDF membranes (EMD Millipore, IPVH00010) at 65 mA. Membranes were rinsed in PBS 2×5 min, fixed in 0.4% paraformaldehyde 2 h and rinsed in PBS five times for 5 min.
Blocking and antibody diluting solutions were buffered with 100 mM Tris (pH 7.5) when blotting for phospho-Smad1/5, Smad1, Smad5, ERK1/2 or phospho-ERK1/2; with 1× TBS when blotting for phospho-Smad2 or Smad2; or with 1× PBS when blotting for Gapdh. Membranes were incubated 1 h in blocking solution [5% chicken serum (Life Technologies, 16110082), 0.1% Tween 20], rinsed 1 min in Tris or PBS and then incubated 1 h in primary antibody solution (PBS or Tris, 3% BSA, 0.2% Tween 20, 0.02% sodium azide) containing: 1:2000 rabbit anti-phospho-Smad1/5 (Cell Signaling, 9516P), 1:2000 rabbit anti-Smad1 (Cell Signaling, 9743P), 1:1000 rabbit anti-Smad5 (Cell Signaling, 9517P), 1:2000 mouse anti-Gapdh (Sigma, G8795), 1:500 rabbit anti-phospho-Smad2 (Cell Signaling, 3108), 1:500 rabbit anti-Smad2 (Cell Signaling, 5339), 1:2000 rabbit anti-phospho-ERK1/2 (Cell Signaling, 4370S) or 1:200 rabbit anti-ERK1/2 (Santa Cruz, sc-94). Membranes were then rinsed three times for 10 min in blocking solution and incubated for 1 h with HRP-conjugated anti-mouse or anti-rabbit antibody (Bio-Rad, 170-6515 and 170-6516) at 1:5000 in blocking solution. They were rinsed four times for 5 min in Tris or PBS with 0.1% Tween 20 and signal was revealed using a Lumi-Light PLUS (Roche, 12 015 196 001) or, for Gapdh, LumiGLO (Cell Signaling, 7003) and auto-radiography films (Progene, 39-20810).
Membranes were used for multiple western blots, between which they were stripped of antibodies. Membranes were incubated in 2% SDS, 100 mM β-mercaptoethanol, 50 mM Tris-HCl (pH 7.5) at 50°C for 15-30 min. They were then thoroughly washed in deionized water, before being reused with another antibody.
Hematoxylin and Eosin, BrdU and TUNEL assays
12 h before harvest, animals were injected intraperitoneally with 10 µl of BrdU solution (Life Technologies, 000103) per gram of body weight. Limbs were harvested and fixed in a large excess of 4% paraformaldehyde in 0.7× PBS, overnight. They were then rinsed three times for 20 min in PBS and embedded in paraffin wax. Paraffin blocks were sectioned at 10 µm and sequential sections were stained with Hematoxylin and Eosin, for BrdU or TUNEL assay. BrdU and TUNEL assays were carried out as described by Guimond et al. (2010), with minor changes. In the BrdU assay, the HRP substrate was CF640R tyramide (Biotium, 92175). For the TUNEL assay, the secondary antibody used was Alexa Fluor 594 goat anti-mouse (Life Technologies, A11020).
Positive cell counting was semi-automated. TUNEL-positive nuclei were counted manually. For DAPI+ and BrdU+ nuclei, channels were taken individually, colors inverted and an auto-contrast was performed with Adobe Photoshop CS6. In ImageJ 1.50i, Subtract Background (Rolling Ball Radius 12 pixel, Light background) was used, then Threshold (auto settings, apply) and Watershed. Nuclei were counted with the Analyse particle function (Size, 120+ pixels, circularity 0.1-1.0).
Paraffin sections were rehydrated and rinsed in PBST. They were then treated with 0.05 µg/ml proteinase K at room temperature for 30 min and rinsed twice for 5 min in PBST. Slides were submerged in HIER solution (10 mM sodium citrate, 1 mM EDTA and 0.1% Tween 20) and heated in a pressure cooker (Cuizen digital pressure cooker) at 7.5 psi for 15 min. They were then rinsed once in HIER solution, twice for 5 min in TBS and once in TBST for 5 min. Sections were incubated with blocking solution (5% chicken serum in TBST) for 1 h, followed by incubation with anti-phospho-Smad1/5 antibody (1:300 in blocking solution) at 4°C overnight. After three 5 min washes in TBST, sections were incubated with HRP-conjugated anti-rabbit antibody (1:400 in blocking solution) for 1 h, then washed again three times for 5 min in TBST. Slides were then rinsed in PBST, incubated for 8 min in CF640R tyramide (11.2 µM in PBST, with 0.015% H2O2) and washed in PBST three times for 5 min. Coverslips were mounted with Prolong Gold (Invitrogen, P36931).
In situ hybridization
RNA probes were prepared using DNA amplicons that were generated by PCR for each gene, with the primers listed in Table S4. The promotor sequence of the T7 RNA polymerase (TAATACGACTCACTATAGGGAGA, not shown in the table) is added to the 5′ of each anti-sense primer. The amplicons were purified with the Purelink PCR purification kit. The probes were synthesized using 9 µl of purified amplicon, 4 µl of T7 5× buffer, 2 µl of 0.1 M DTT, 1 µl of RNAseOUT, 2 µl of 10× RNA DIG labeled NTP mix (Roche 11277073910) and 2 µl of T7 RNA polymerase (Invitrogen 18033019). The solution was incubated at 37°C for 3 h. The template DNA was digested by adding 4 µl of DEPC-treated water, 3 µl of 10× DNAseI buffer, 1 µl of RNAseOUT and 2 µl of DNAseI (Invitrogen 18068015), then incubating at 37°C for 30 min. The probes were precipitated by adding 4 µl of glycogen 5 µg/µl, 2.5 µl of 4 M LiCl, 1 µl of 0.5 M EDTA and 82 µl of 100% ethanol, then incubating at −20°C overnight. The RNA was pelleted and washed with 500 µl of 75% ethanol. It was then dissolved in 50 µl of DEPC-treated water, quantified and diluted at 10 µg/ml in hybridization solution (4× SSC, 50% formamide, 1 mg/ml Torula yeast RNA, 0.1 mg/ml heparin, 1× Denhart's solution, 0.1% Tween 20, 0.05% CHAPS and 5 mM EDTA).
Paraffin sections were re-hydrated and rinsed in PBST (made with DEPC-treated water). The sections were treated with proteinase K (in PBST, see Table S4 for concentration) at room temperature for 30 min and rinsed twice for 5 min in PBST. They were refixed in 1% PFA (in PBST) at room temperature for 30 min and rinsed twice for 5 min in PBST. They were treated with 0.25% acetic anhydride (in 0.1 M triethanolamine), at room temperature for 10 min and rinsed twice for 5 min in PBST. Sections were incubated in hybridization solution at the hybridization temperature (Hyb To, see Table S4) for 1 h. The probe was diluted at 1 µg/ml in hybridization solution containing 10% dextran sulfate and denatured at 80°C for 5 min. Sections were then incubated with the probe at the Hyb To, covered with a coverslip and in a humid chamber, for 16 h.
On the second day, sections were rinsed in 2× SSC at Hyb To for 10 min, and incubated in hybridization solution at Hyb To, 30 min. They are then washed at Hyb To twice in 2× SSC for 10 min, followed by two washes in 0.2× SSC for 10 min. Sections were washed in PBS at room temperature for 5 min. The sections were incubated in the blocking solution (10% sheep serum in MABT) for 1 h. They were then incubated with 2 µg/ml anti-Dig antibody (Roche 11333062910, in blocking solution) at 4°C overnight.
On the third day, sections were washed in MABT, three times for 5 min, and incubated with HRP-conjugated anti-mouse antibody (1:400 in blocking solution) at room temperature for 1 h. They were washed in MABT, three times for 5 min, then in PBST for 5 min. Sections were incubated for 8 min with CF640R tyramide (11.2 µM in PBST with 0.015% H2O2) and washed in PBST three times for 5 min. Coverslips were mounted with Prolong Gold.
Injections of FGF8 protein
FGF8 protein (Human FGF8 from R&D Systems, 424-FC/CF; Axolotl FGF8 produce in-house by overexpressing the protein, containing a C-terminal His-tag, in E. coli) was dissolved in 0.7× PBS at 0.5 µg/µl. Animals with both forelimbs amputated were treated with DMSO or 0.5 µM LDN193189 from day 5 to day 12 post-amputation. Once a day, from day 7 to 11, the right limb blastema was injected with 0.3 µl of either PBS or FGF8 solution (with 0.01% Fast Green). Injection needles were made using borosilicate capillaries (Sutter Instrument B100-75-10), with the inside coated with Sigmacote (Sigma, SL-2). They were pulled with a micropipette puller (Sutter Instrument P-87) and the tip of the needles were broken to have a sharp tip end with a diameter of 20 µm. Needles were connected to a pneumatic injector (MicroData Instruments, PM 1000 Cell microinjector) and the pressure was adjusted to slightly above what is required to eject deionized water. The solutions were then aspirated and quickly injected into the blastema. Needles were changed after each injection.
Pictures and image editing
Pictures of live animals and whole limbs were taken on a Leica MZ16F stereomicroscope, using a Lumenera Infinity 2 camera. Pictures of sections were taken using a Zeiss Axio Imager M2 microscope and Axiocam 506 color or Axiocam MRm monochrome camera. Images were recorded and stitched together with the Zeiss Zen 2 Pro program. Cropping and level adjustments were made using Adobe Photoshop CS6. Drawings of regenerating limbs were made by tracing pictures of regenerating limbs with Adobe Illustrator CS6. Some modifications were made to the tracings. Figures were assembled with Adobe Illustrator CS6.
We acknowledge the help of Dr Antonio Nanci for use of histology material and microscopes in his lab.
Conceptualization: E. Vincent, S.R.; Methodology: E. Vincent, E. Villiard, F.S., S.D., B.K., S.R.; Validation: E. Vincent, E. Villiard; Formal analysis: E. Vincent; Investigation: E. Vincent, S.R.; Writing - original draft: E. Vincent, S.R.; Writing - review & editing: E. Vincent, E. Villiard, F.S., S.R.; Visualization: E. Vincent; Supervision: S.R.; Funding acquisition: S.R.
This research is supported by a grant from the Canadian Institutes of Health Research (MOP: 111013 to S.R.).
The complete axolotl sequence has been deposited in Genbank under accession number KX809594.
The authors declare no competing or financial interests.