Adrenergic signaling coordinates distant and local responses to amputation in axolotl

Abstract

SUMMARY

Many species regenerate lost body parts following amputation. Most limb regeneration research has focused on the immediate injury site. Meanwhile, body-wide injury responses remain largely unexplored but may be critical for regeneration. Here, we discovered a role for the sympathetic nervous system in stimulating a body-wide stem cell activation response to amputation that drives enhanced limb regeneration in axolotls. This response is mediated by adrenergic signaling, which coordinates distant cellular activation responses via the α_2A-adrenergic receptor, and local regeneration responses via β-adrenergic receptors. Both α_2A- and β-adrenergic signaling act upstream of mTOR signaling. Notably, systemically-activated axolotls regenerate limbs faster than naïve animals, suggesting a potential selective advantage in environments where injury from cannibalism or predation is common. This work challenges the predominant view that cellular responses underlying regeneration are confined to the injury site and argues instead for body-wide cellular priming as a foundational step that enables localized tissue regrowth.

IN BRIEF

The nervous system regulates a systemic response following amputation that primes distant appendages for faster regeneration upon a second injury, a scenario relevant to salamanders in the wild who experience repeated limb loss.

Graphical Abstract

INTRODUCTION

The ability to spontaneously replace large body parts lost to injury or disease varies across the animal kingdom. Salamanders, such as axolotls, can regenerate full limbs, but mammals cannot. Relatively little is known about how selective pressures experienced by salamanders might have shaped their remarkable regenerative traits. Furthermore, despite progress in understanding the molecular responses of salamander cells at the amputation site, how cells elsewhere in the body respond to amputation of a distant limb and the implications of these responses have been less explored. Yet, these distant responses could be important for understanding limb regeneration and the evolution of this trait, and they could provide essential insights for future regenerative medicine approaches aimed at harnessing latent regenerative capacity in humans.

Previously, we found that in axolotls, limb amputation provokes a body-wide cellular proliferation response within many distant tissues in a process we called “systemic activation.”¹ Systemic activation occurs even when limbs are experimentally blocked from regenerating, indicating it is not likely to be driven by growth factors secreted by regenerating limbs. However, the molecular signals that stimulate systemic activation remained unknown. In this earlier work, we also found that an ancient pathway controlling cell growth across many species, mTOR, is enhanced both at the amputation site and within distant, responding tissues. However, whether mTOR signaling is required for systemic activation in axolotls, as well as the potential upstream signals that might regulate mTOR in this context, remained unknown. A further mystery has been whether the signaling pathways that govern systemic activation are unique to distant locations, or whether they might also drive cellular proliferation and blastema formation at the amputation site. We therefore sought to uncover how distant cells are cued by amputation to become systemically activated in axolotl, the relationship of these stimuli to mTOR signaling, and their potential to promote blastema formation at the injury site. We also aimed to uncover the implications of systemic activation on axolotls.

Here, we show that systemic activation primes distant limbs for faster regeneration in axolotls, a response that may be evolutionarily relevant in wild salamander populations. We demonstrate that peripheral innervation of both the amputation site and the distant, responding site is required for systemic activation. We further find that the sympathetic nervous system mediates this dependency through α-adrenergic signaling. We also show that mTOR signaling operates downstream of α-adrenergic signaling in this process. Interestingly, while α-adrenergic signaling is required for systemic activation, β-adrenergic signaling is dispensable. However, β-adrenergic signaling is necessary at the amputation site for limb regeneration, where it promotes mTOR signaling, cellular proliferation, and blastema formation. Our work demonstrates that a common signal, the adrenergic ligand norepinephrine, is interpreted by cells differently depending on their proximity to the amputation site. In the amputated limb tissues, both α_2A- and β-adrenergic receptor activities promote mTOR signaling and proliferation by blastemal progenitors, while at distant sites elsewhere in the body, α-adrenergic receptors stimulate mTOR to promote systemic activation, thereby priming axolotls for enhanced regenerative responses to future injuries. These results highlight potentially evolutionarily relevant responses to amputation that could be explored in other species to promote regenerative therapies.

RESULTS

Amputation primes distant uninjured limbs for future local regeneration

Recently, we uncovered a phenomenon of body-wide cellular proliferation response to amputation in axolotl (Ambystoma mexicanum). We sought to understand the consequences of systemic activation to axolotls. We first asked whether systemically-activated axolotls are more, or less, effective at mounting future regenerative responses. We challenged systemically-activated axolotls with amputation (Figure 1A) and found they regenerate these limbs, distant to the originally-amputated limbs, significantly faster than naïve control animals (Figure 1B). Systemically-activated limbs grew blastemas more quickly (Figure 1C) and reached digit-differentiation stages faster (Figure 1D-E). We asked whether this enhanced regeneration response is facilitated by quicker blastema cell specification and coalescence. We visualized the expression of an established blastema transcript, Prrx1, and found that systemically-activated limbs have more robust blastema-initiation responses than naïve limbs (Figure 1F-G). This data supports the conclusion that amputation-induced systemic activation leads to priming of distant tissues for faster regeneration in animals subjected to future local injury. A faster recovery of digits, for example, may be critical for animal fitness and survival in the wild.

Figure 1. Systemically-activated cells are primed for future regeneration events.

(A) Regeneration rates of systemically-activated and naïve limbs. Green dots, EdU+ nuclei. (B) Regenerating systemically-activated or control limbs (ventral view). (C) Blastema area at 14 dpa (n=15 per group), (D-E) Number of differentiating limbs (D) and digit outlines (E) at 25 dpa (n=30 limbs/group). (F) Prrx1 HCR-FISH of regenerating naïve and primed limbs at 5 dpa, quantified in (G) (n=8 per group). (H) EdU staining of homeostatic and contralateral limbs harvested 6 weeks post-EdU administration, quantified in (I) (control n=15, activated n=12). (J) Systemically-activated and naïve limbs amputated 28 days post-contralateral amputation (ventral view, 7–21 dpa; dorsal view, 27 dpa). (K) Blastema area at 7 dpa (control n=19, activated n=20). (L) Number of digit outlines at 27 dpa (control n=38 limbs, activated n=40 limbs). (M) Tiger salamander simultaneously regenerating three limbs. Data shown as mean ± SD. Statistical significance determined using unpaired two-tailed Welch’s t-test in (C), (G), (I) and (K); and Fisher’s exact test in (D), and Fisher’s exact test with Freeman-Halton extension in (E) and (L). Scale bars, 2 mm in (B) and (J); 500 μm in (F) and (H).

We investigated how many divisions cells undergo following their systemic activation and how long the priming effect lasts. Using a short-pulse/long chase EdU strategy, we found most EdU-labeled clones in limbs are composed of 2–8 cells (Figure 1H), indicating that amputation-induced proliferation lasts only a few cell cycles. Concordantly, we found no difference in the fraction of EdU+ cells by 6 weeks post-amputation (Figure 1I). Six weeks is enough time for larval axolotls to completely regenerate a limb. We also found that systemically-activated limbs challenged with local amputation 4 weeks after initial activation regenerated at the same rate as naïve limbs, indicating the effect stimulated by initial limb loss is relatively short-lived (Figure 1J-L).

We considered whether there could be a need for wild salamanders to sequentially regenerate multiple limbs over a short timeframe. This is conceivable, as limb and tail regeneration following conspecific or predatory bite wounds is frequent in wild salamander populations^2,3; salamander species even use a form of tail autotomy.³ Axolotls are critically endangered in the wild⁴ and could not be investigated in a natural setting. We therefore examined tiger salamanders (Ambystoma tigrinum), close cousins of axolotls, found in a naturalistic outdoor pond setting, for evidence of simultaneous multiple limb regeneration. We found that many larval tiger salamanders had multiple limbs at different, off-set states of regeneration in this natural setting (Figure 1M; N>6), implying that the amputation injuries occurred sequentially rather than simultaneously. The loss of a first limb could reduce mobility and predispose an animal to subsequent predation. Loss of multiple limbs in individuals has been documented in wild salamander populations, such as eastern hellbenders (Cryptobranchus alleganiensis alleganiensis)⁵ and ocoee salamanders (Desmognathus ocoee).⁶ These data indicate that in certain settings, limb loss is likely to occur repetitively and is physiologically relevant for salamanders. Collectively, these data demonstrate that prior limb loss can enhance new limb regeneration elsewhere within the period of systemic activation, suggesting relevance to repeated limb loss in the wild.

Sympathetic nervous system is required for systemic activation and priming for limb regeneration

Understanding how amputation signals spread is key to uncovering the molecular and cellular basis of systemic activation. We hypothesized that the peripheral nervous system (PNS) might be a conduit for this type of information, as it rapidly transduces sensory afferent information to the spinal cord and brain, as well as sends motor and autonomic efferent signals from the brain to the body, thereby facilitating communication between distant anatomical locations. We tested whether systemic activation requires peripheral innervation at either the responding site (Figure 2A-E) and/or the amputation site (Figure 2F-G) using surgical denervation via brachial plexus nerve transection or sciatic nerve transection and comparing to sham-operation controls (Figure S1). We found that systemic activation did not occur when these distant sites contralateral to amputation were denervated at the time of the amputation injury (Figure 2A-B). As a control, we found that systemic activation still occurred following amputation when these distant sites (contralateral limbs) underwent sham operations. We parsed the contralateral tissues into three broad tissue categories—epidermis, skeletal elements, and internal soft tissues—and quantified the extent of systemic activation within each (Figure 2C-E). This analysis revealed that each of these three tissue types requires an intact nervous system for cellular activation. We also found that when limbs were denervated and then immediately amputated, the amputation no longer provoked systemic activation in contralateral limbs (Figure 2F-G). Sham surgeries still provoked the systemic activation response. These results demonstrate that peripheral innervation of both the amputated and the distantly-responding limb is required for systemic activation. Therefore, both neuronal input that senses injury and neuronal output that mediates efferent signaling to tissues could regulate systemic activation.

Figure 2. PNS at both amputation site and responding site is required for systemic activation.

(A-E) Nerve bundles innervating limbs contralateral to amputations were transected or sham-operated. (A) EdU staining of intact and contralateral limbs of denervated and sham-operated animals in the intact state or at 14 dpa, quantified in (B). (C-E) Breakdown of tissue types represented in (B). (C) Epidermis. (D) Skeletal elements. (E) Internal soft tissues (sham n=11, denervated n=10, sham amputated and denervated amputated n=13). (F-G) Nerve bundles were transected or sham-operated in limbs that were immediately amputated thereafter. (F) EdU staining of intact and contralateral limbs of denervated and sham-operated animals in the intact state or at 14 dpa, quantified in (G) (sham n=18, sham amputated n=16, denervated n=11, denervated amputated n=12). (H) Sympathetic nerves and EdU+ cells in adjacent tissues in homeostatic and systemically-activated limbs. (I) TH staining of vehicle or 6-OHDA-treated limbs. (J) EdU staining of 6-OHDA or vehicle-treated contralateral limbs at 14 dpa, quantified in (K) (vehicle n=21, 6-OHDA n=20). (L) Regenerating limbs following 6-OHDA or vehicle treatment (ventral view). (M) Blastema area at 14 dpa (vehicle n=21, 6-OHDA n=20), (N) EdU staining of 6-OHDA or vehicle-treated limbs at 14 dpa, quantified in (O) (vehicle n=21, 6-OHDA n=20). Data shown as mean ± SD. Statistical significance determined using unpaired two-tailed Welch’s t-tests with Bonferroni-Dunn correction in (B), (C), (D), (E) and (G), unpaired two-tailed Welch’s t-tests in (K), (M) and (O). Scale bars, 500 μm in (A), (F), (J) and (N), 200 μm in (H), 100 μm in (I), and 2 mm in (L). den, denervated. int, intact. contra, contralateral. See also Figure S1.

We next addressed which component of the peripheral nervous system regulates systemic activation. We hypothesized that limb loss is likely to provoke canonical stress pathways, such as activation of the sympathetic nervous system mediating the “fight or flight” response. Supporting a possible role for sympathetic innervation, we noticed in our tissue sections that EdU+ cells were located near sympathetic nerves (Figure 2H). We therefore tested whether sympathetic innervation is required for systemic activation. We intraperitoneally administered 6hydroxydopamine (6-OHDA) to axolotls to ablate sympathetic nerve processes and confirmed this effect in the periphery, in limb tissues (Figure 2I). Following 6-OHDA treatment, we amputated limbs and assayed for systemic activation. We found that sympathetic nerve ablation blocks systemic activation (Figure 2J-K). Intriguingly, we also found that amputated limbs displayed significant impairment in blastema formation in the 6-OHDA-treated animals (Figure 2L-M). Furthermore, these 6-OHDA-treated limbs contained significantly fewer proliferating cells than vehicle controls (Figure 2N-O). These results demonstrate that sympathetic innervation is required for both systemic activation and for local limb regeneration. They also reveal mechanistic links between systemic limb cell responses to amputation and their role in local regeneration. We therefore sought to identify the molecular effectors of noradrenergic signals in systemic activation and limb regeneration.

α_2A adrenergic receptor activity is required for systemic activation and priming

The primary adrenergic signaling effector of the sympathetic nervous system is noradrenaline, which acts as a neurotransmitter to provoke responses in peripheral tissues.⁷ Noradrenaline can bind to several different adrenergic receptors. Pharmacological inhibitors exist that antagonize specific noradrenaline-receptor interactions. We used these to address which adrenergic receptors are relevant to systemic activation and limb regeneration. We first treated axolotls with yohimbine, a well-studied ADRA2A antagonist,^8,9 to test for an adrenergic signaling requirement in both the regenerating limb and in systemic activation. In yohimbine-treated animals, amputated limbs had lower numbers of proliferating cells (Figure 3A-B), and limbs contralateral to amputation did not undergo systemic activation (Figure 3C-D), demonstrating that adrenergic signaling is required for robust cell proliferation at the injury site and for systemic cell activation in response to amputation. We then asked whether adrenergic signaling is also required for priming distant tissues for future regeneration (Figure 3E). We found that when axolotls were treated with yohimbine from 2 days before to 7 days after amputation, priming was diminished (Figure 3F), with both initial blastema growth (Figure 3G) and limb differentiation (Figure 3H) occurring more slowly. We tested whether stimulating adrenergic signaling via the α_2A-adrenergic agonist clonidine might enhance priming and found evidence that this pathway can indeed be stimulated to accelerate blastema growth of distant tissues that have been systemically activated (Figure 3I-K). Collectively, these data support a model in which noradrenaline released from sympathetic neurons provokes systemic activation through α_2A-adrenergic receptors in target tissues, that this process is necessary for priming cells for faster future regeneration, and that targeting this pathway with a pharmacological agonist can enhance regenerative priming.

Figure 3. α-Adrenergic signaling is required for systemic activation and priming.

(A) EdU staining of yohimbine or vehicle-treated blastemas at 12 dpa, quantified in (B) (vehicle n=20, yohimbine n=13). (C) EdU staining of yohimbine or vehicle-treated contralateral limbs at 7 dpa, quantified in (D) (n=15 per group). (E) Experimental design for (F-K). (F) Regenerating contralateral limbs of yohimbine and vehicle-treated animals (ventral view, 5–25 dpa; dorsal view, 32 dpa), blastema area at 14 dpa quantified in (G) (n=15 per group), number of digit outlines at 25 dpa quantified in (H) (n=30 limbs/group). (I) Regenerating contralateral limbs of clonidine and vehicle-treated animals (ventral view, 7–23 dpa; dorsal view, 33 dpa), blastema area at 7 dpa quantified in (J) (vehicle n=20, clonidine n=18), number of digit outlines at 23 dpa quantified in (K) (vehicle n=40 limbs, clonidine n=35 limbs). The same group of vehicle-treated animals was used in Figure 4 (I-K). (L) EdU staining of propranolol or vehicle-treated blastemas and contralateral limbs at 7 dpa. Quantification of EdU signal in blastema (M) and contralateral limbs (N) (vehicle n=15, propranolol n=10). (O) Clonidine or vehicle-treated blastemas at 7 dpa, quantified in (P) (vehicle n=25, clonidine n=20). (Q) EdU staining of regenerating limbs 7 dpa, quantified in (R) (vehicle n=15, clonidine n=14). Data shown as mean ± SD. Statistical significance was determined unpaired two-tailed Welch’s t-tests in (B), (D), (G), (M), (N), (P) and (R); Welch’s ANOVA with Dunnett’s T3 correction in (J); Fisher’s exact test with Freeman-Halton extension in (H) and (K). Scale bars, 500 μm in (A), (C), (L) and (Q); 2 mm in (F), (I) and (O). See also Figure S2.

We assessed possible contributions to systemic activation from other adrenergic receptors that can also bind noradrenaline. We used an identical treatment regimen to administer the non-selective β-adrenergic receptor antagonist propranolol¹⁰ to axolotls (Figure 3L-O). We found that while propranolol treatment led to a significantly decreased fraction of cells in S-phase within blastemas of regenerating limbs (Figure 3L-M), it had no effect on systemic activation (Figure 3L, N). These results indicate β-adrenergic signaling is only required for the localized cellular proliferation response at the amputation site.

We next asked whether stimulating α_2A-adrenergic signaling in the context of a single limb amputation alone is sufficient to promote faster blastema formation, and, using clonidine treatment, we found that this was the case (Figure 3O-P). This enhanced regeneration response is accompanied by increased proliferation in the amputated limbs (Figure 3Q-R). We also tested whether clonidine administration is sufficient to promote systemic activation in the absence of amputation injury and found it to be insufficient (Figure S2B-C). Clonidine treatment also could not rescue systemic activation when either amputated limbs (Figure S2D-E) or responding limbs (Figure S2F-G) were denervated. Together, these results argue that systemic activation and priming require α-adrenergic signaling, and that stimulating this pathway is sufficient to promote faster limb regeneration, likely synergizing with other pathways downstream of nerves. These findings also argue that β-adrenergic signaling is required specifically at the injury site to promote limb regeneration.

mTOR signaling is required for systemic activation in axolotl

Our findings highlight the sympathetic nervous system and adrenergic signaling as key drivers of systemic priming and proliferation after amputation in axolotls. A similar systemic response occurs in mice after local injury whereby distant cells re-enter the cell cycle.¹¹ In mice, this activation is linked to circulating HGFA (hepatocyte growth factor activator), which activates HGF (hepatocyte growth factor) and stimulates mTOR signaling.^11,12 Arecent study of the amputation site identified a key role for mTOR in translational regulation of the wound-healing stage of axolotl limb regeneration.¹³ We investigated the potential participation of mTOR in systemic activation in axolotl. We found that axolotls treated with the mTOR inhibitor rapamycin¹⁴ regenerate limbs more slowly (Figure 4A-C), with fewer proliferating cells in the blastema (Figure 4D-E), and that both systemic activation (Figure 4F-G) and priming are inhibited by rapamycin treatment (Figure 4H-K). Considering the role of mTOR in nutrient sensation, cell growth, and proliferation,¹⁵ we tested whether fasted axolotls exhibit altered levels of systemic activation and whether cell proliferation in this context might additionally be regulated by mTOR signaling. We found that global cell proliferation rates in amputated and fasted axolotls were dramatically lower than typically observed in well-fed animals (Figure 4L-M). Yet, we also found that in the context of an amputation, the number of proliferating cells at distant sites still reduced by inhibition of mTOR signaling (Figure 4L-M). These results demonstrate that mTOR signaling is important for systemic activation responses to limb amputation in axolotl.

Figure 4. mTOR signaling is required for systemic activation and priming.

(A) Rapamycin or vehicle-treated regenerating limbs at 7 dpa, quantified in (B) (n=10 per group). (C) Alcian blue and Alizarin red staining at 10 weeks post-amputation. (D) EdU staining of rapamycin or vehicle-treated blastemas at 14 dpa, quantified in (E) (vehicle n=5, rapamycin n=4). (F) pS6 and EdU staining of saline, vehicle or rapamycin-treated homeostatic and contralateral limbs at 14 dpa, EdU quantified in (G) (intact saline, intact vehicle and intact rapamycin n=12, amputated saline and amputated vehicle n=10 limbs, and amputated rapamycin n=8). (H) Experimental design to test for mTOR signaling in priming. (I) Regenerating contralateral limbs of rapamycin and vehicle-treated animals, blastema area at 7 dpa quantified in (J) (n=20 per group), number of digit outlines at 23 dpa quantified in (K) (n=40 limbs/group). The same group of vehicle-treated animals were used in Figure 3 (I-K). (L-M) Animals fasting for four weeks were treated with rapamycin or vehicle solution. (L) EdU staining of rapamycin or vehicle-treated contralateral limbs at 14 dpa, quantified in (M) (n=14 per group). Data shown as mean ± SD. Statistical significance determined using unpaired two-tailed Welch’s t-test with Holm-Šídák correction in (B); unpaired two-tailed Welch’s t-test in (E) and (M); unpaired two-tailed Welch’s t-test with Bonferroni-Dunn correction in (G); two-tailed Welch’s ANOVA with Dunnett’s T3 correction for multiple hypothesis testing in (J); and Fisher’s exact test with Freeman-Halton extension in (K). Scale bars, 0.5 mm in (A); 500 μm in (D), (F) and (L); 2 mm in (I). Arrowheads mark the amputation plane. rapa, rapamycin. Int, intact. contra, contralateral.

Adrenergic signaling operates upstream of mTOR signaling in systemic activation and limb regeneration

We next asked whether mTOR and adrenergic signaling requirements in systemic activation are mechanistically linked. We inhibited α_2A-adrenergic signaling using yohimbine, amputated limbs, and assayed for mTOR signaling in contralateral limbs using pS6. We found that yohimbine treatment significantly reduced mTOR signaling in contralateral limbs, providing evidence that mTOR is under the control of α_2A-adrenergic signaling in the systemic response to amputation (Figure 5A-B). In separate experiments, we treated naïve axolotls with the yohimbine or with the β-adrenergic antagonist propranolol, amputated limbs, and assayed for mTOR signaling in each case using pS6 as a readout. We found that both yohimbine treatment (Figure 5C-D) and propranolol treatment (Figure 5E-F) significantly reduced mTOR signaling in stump tissues. These data argue that adrenergic signaling is required for mTOR signaling in the context of axolotl limb regeneration. To further investigate mTOR signaling dependency on adrenergic signaling, we leveraged an existing axolotl limb fibroblast cell line, AL-1 cells.¹⁶ We treated AL-1 cells with adrenaline and found it was sufficient to significantly enhance mTOR signaling in vitro (Figure 5G-H). Collectively, these data argue that axolotl limb cells use adrenergic signals to upregulate mTOR signaling.

Figure 5. Adrenergic signaling operates upstream of mTOR signaling in systemic activation and limb regeneration.

(A) pS6 staining of yohimbine or vehicle-treated contralateral limbs at 12 dpa, quantified in (B) (vehicle n=13, yohimbine n=15). (C) pS6 staining of yohimbine or vehicle-treated regenerating limbs at 12 dpa, quantified in (C) (vehicle n=18, yohimbine n=11). (E) pS6 staining of propranolol and vehicle-treated regenerating limbs at 7 dpa, quantified in (F) (vehicle n=15, propranolol n=11). (G) pS6 staining of AL-1 cells treated with adrenaline, rapamycin or vehicle solution for 24 hours, quantified in (H) (n=3 per group). Data shown as mean ± SD. Statistical significance was determined using unpaired two-tailed Welch’s t-tests in (B), (D) and (F); ordinary one-way ANOVA with Dunn-Šidák correction for multiple hypothesis testing in (H). Scale bars, 500 μm in (A), (C) and (E); 50 μm in (G).

Systemically-activated cell types are shared with proliferating cell types in homeostasis and are epigenetically distinct

We sought to identify the types of cells that become systemically activated and the signals they might use to communicate with other cells in activated limbs. In our earlier study, we identified a subset of systemically-activated cells (SACs) as satellite cells¹, which serve as stem cells for muscle regeneration in many species, including axolotls.¹⁷ However, the identities of most axolotl SACs remained unknown. We designed a marker-agnostic FACS strategy to purify and profile the transcriptomes of individual SACs based on their 4C nuclear content compared to 2C content of non-proliferative cells (Figure 6A-B). We profiled transcriptomes from 4C SACs and 2C cells from both systemically-activated and naïve control limbs. We merged these data and classified cell types across samples (Figure 6C, Figure S3, Table S1, Table S2). The vast majority of proliferative 4C cell types had correlating clusters among non-proliferative 2C cells (Figure 6D). Furthermore, the SAC types overlap with 4C cell types collected from intact, naïve limbs (Figure 6E). These data indicate that SACs largely represent stem and progenitor cell types that ordinarily cycle during homeostasis, including during organismal growth and wear-and-tear repair. This finding suggests that regenerative strategies could be developed to potentially steer the activities of pre-existing progenitor cell types from growth and homeostatic functions toward building a blastema including, conceivably, in mammals.

Figure 6. Systemically-activated cells represent the same broad cell types as cells proliferating during homeostasis and are epigenetically primed.

(A) Experimental design. (B) FACS distribution of homeostatic and systemically-activated limb cells. (C) UMAP homeostatic and systemically-activated limb cells (n=4, each). (D) UMAP of 2C cells in red and 4C cells in blue. (E) UMAP of 4C cells from the homeostatic limb in blue and systemically-activated limb in yellow. (F) Expression of blastema-specific genes across Adra2a+ and Adra2a-cells in homeostatic (H) versus activated (A) samples. (G) Alpha- and beta-adrenergic receptor expression in regenerating limbs (7 timepoints from 3 hpa to 33 dpa combined), data re-analyzed from Li et al.⁴⁸ (H) Marco HCR-FISH in propranolol or vehicle-treated blastemas at 7 dpa, quantified in (I) (vehicle n=12, propranolol n=11). (J) Significantly enriched transcription factor binding motifs within open chromatin of 4C cells. Representative factors are noted; green shows factors directly associated with EMT. (K) EMT marker expression in homeostatic and activated fibroblasts. (L) Kazald2 and Snai1 HCR-FISH on homeostatic and activated limbs. (M) Kazald2 and Snai1 RNAscope-FISH on wholemount homeostatic limbs. Data shown as mean ± SD. Statistical significance determined using unpaired two-tailed Welch’s t-test in (I). Scale bars, 500 μm in (H) and (L). See also Figure S3, S4 and S5, Table S1, S2 and S3 and Supplementary Movie 1.

We also used these transcriptome data to identify cell types that express adrenergic receptors. We found that in both homeostatic and activated limbs, fibroblasts are the primary cell type expressing Adra2a transcripts (Figure S4A), while β-adrenoreceptor transcripts are expressed by lymphocytes, mast cells, macrophages and endothelial cells in both homeostatic and activated limbs (Figure S4B). This data supports a model in which limb fibroblasts receive norepinephrine signals, potentially acting as sensors of distant amputation. We wondered if cells that express Adra2a might be primed toward a more regenerative state compared to other limb cells. We therefore computationally isolated Adra2a⁺ cells from our dataset and performed co-expression analyses with established blastema-enriched genes. We found that Adra2a⁺ cells in both homeostatic and activated limbs are enriched for expression of a suite of such genes, including Prrx1, Prrx2, Vwde, Kazald2, and Twist1, among others (Figure 6F). These data indicate that Adra2a-expressing cells are likely primed toward blastema cell states even before distant injury and that distant amputation provokes upregulation of genes that typically rise during bona fide blastema creation on amputated limb stumps.

We next sought to relate our findings about adrenoreceptor expression to cellular events regulating localized blastema formation within amputated limb stumps. We interrogated a published single-cell RNA-sequencing dataset¹⁸ that profiled cells from homeostatic limbs and from limb stumps post-amputation. Intriguingly, we found the most significantly expressed adrenoreceptor to be Adrb2, which was expressed in several types of immune cells, including macrophages, as well as in epithelial cell populations of wound epidermis (Figure 6G). These results reinforce our earlier experimental finding that β-adrenergic signaling is important for local blastema formation and limb regeneration even though it is dispensable for systemic activation. Macrophages are known to be required at the amputation site for successful blastema formation and limb regeneration¹⁹, but the upstream signals that recruit them to the injury site are largely unknown. We hypothesized that adrenergic signaling and specifically epinephrine-mediated signaling through β-adrenoreceptors may be required for macrophage recruitment at the amputation site. We blocked β-adrenergic signaling using propranolol and assayed for macrophages at the amputation site using HCR RNA in situ hybridization for the macrophage marker Marco (Figure 6H). We discovered that propranolol-treated samples harbored significantly fewer Marco⁺ macrophages at the amputation site (Figure 6I). A previous report demonstrated that immune cytokine IL-8 can promote recruitment of myeloid cells to regenerating axolotl limbs²⁰, and in human monocytes, β-adrenergic signaling in myeloid cells can increase IL-8 release.²¹ These results support a model whereby β-adrenergic signaling promotes limb regeneration by controlling macrophage recruitment and/or proliferation during the blastema-forming stages of regeneration.

While differences in gene expression could be important, epigenetic differences may also contribute to systemic activation and the behaviors of SACs. To explore the chromatin accessibility profiles of SACs, we performed bulk ATAC-seq on SACs and 4C cells from homeostatic limbs, as well as blastema tissues. Intriguingly, we uncovered 95 transcription factor binding motifs enriched in regions of open chromatin in both SACs and blastema cells but not in 4C cells from homeostatic, naïve axolotls (Figure 6J, Table S3). Among the open chromatin regions in 4C cells, we found several containing predicted binding motifs for transcription factors with direct roles in regulating epithelial-mesenchymal transition (EMT) in other contexts, such as Twist, Egr1, Etv1, Tcf12, Tcf21, Spdef, Smad3, Jun and AP-1. EMT-like processes are involved in wound healing, blastema formation, cell migration and tissue re-patterning during limb regeneration.^22,23 Additionally, in other contexts, mTOR signaling interacts with EMT processes.²⁴ We reasoned that epigenetic regulation of these pathways in SACs could be important in priming them for future local regeneration. To investigate this possibility, we leveraged our scRNA-seq of homeostatic and systemically-activated limbs. We focused on fibroblasts, as they are an important source of many limb blastema cells.^25–28 We found that several EMT-related transcription factors were upregulated in fibroblasts of the systemically-activated limb compared to the homeostatic limb (Figure 6K). Among these were Egr1, Twist1, Snai1, Snai2 and Zeb1. In our transcription factor binding site analysis, we also found that Egr1 binding sites were shared across homeostatic, activated and regenerating limbs, whereas Twist binding sites were unique to activated limbs (Figure 6J).

To visualize the expression of EMT regulators with respect to proliferating cells during homeostasis and systemic activation, we detected expression of Snai1. In our dataset, Snai1 is upregulated in the Fibroblast I cluster upon systemic activation (Figure 6K). We co-localized Snai1 signal with Kazald2, a unique marker of this fibroblast cluster, known for its regulation of limb regeneration.²⁹ We found that in both homeostatic and activated samples (Figure 6L), some cells localizing to cartilage, perichondrium, nerve bundles, and muscle express Kazald2, Snai1, or both. In homeostatic samples, we observed the strongest co-expression signal for Kazald2 and Snai1 in the peri-skeletal tissue region (Figure 6L). Upon distant amputation injury, we observed a marked increase in expression of both Kazald2 and Snai1 (Figure 6L). We further looked at localization of Kazald2 and Snai1 transcripts in a wholemount limb and confirmed strong co-expression of the two genes in the epidermis and peri-skeletal tissues (Figure 6M, Figure S5 and Supplementary Movie 1). The findings suggest that EMT-like processes may be initiated at the epigenetic and transcriptional level by systemic activation. Interestingly, when we performed a short EdU pulse followed by a long chase (Figure 1H), we found that the resulting 2–8 cell clones in the systemically-activated limb remained clustered within their niches, suggesting that additional mechanisms restrict EMT-like progression until a local injury cue is present.

Collectively, our data demonstrate global differences in chromatin accessibility between homeostatic and systemically-activated states. They further provide evidence that SACs in uninjured tissues bear molecular resemblance to bona fide blastema cells, which may form a cellular basis for faster response in systemically-activated animals to future local injuries.

DISCUSSION

Our results highlight a previously-unappreciated role for body-wide activation of tissue-resident cell types that cycle during organismal homeostasis and their repurposing for pro-regenerative responses during vertebrate limb regeneration. Our work indicates that the initiation phase of limb regeneration engages much more of the animal than previously appreciated and consists of multiple steps that are separable in both space and time. Most previous work investigating regeneration initiation focused exclusively on the local injury site. However, emerging new data shows that during axolotl tail regeneration, immediate early response genes are activated in the brain³⁰, and in developing axolotl embryos, stimulating the brain accelerates tail bud regeneration.³¹ We previously demonstrated that axolotls respond to limb amputation by systemically activating a subset of cells in distant tissues to re-enter the cell cycle but did not address the mechanism of the response or its role in limb regeneration. The present study demonstrates that systemic activation is a priming step in the regenerative process on which subsequent cell behaviors necessary for blastema creation and, ultimately, regeneration are built. These results indicate an EMT-like event is likely the priming mechanism in distant tissues. This EMT event could tip cells toward regeneration by enabling their faster mobilization out of tissue niches after future local injury. These findings indicate that connecting global events to local signaling and cell behaviors is crucial to understanding the regeneration process.

We have here established that systemic stem cell activation is a key feature of the early regenerative response, highlighting the importance of understanding how this process is stimulated by amputation. These data support a model in which the mechanisms of systemic activation require peripheral innervation at both the injury and distant responding sites (Figure 7A) and define a primary role for sympathetic innervation in the initial stimulation of progenitor cells to re-enter the cell cycle. This role for the PNS is complementary to the previously-appreciated roles for the PNS as a source of mitogens in growing nascent blastemas larger^32–36 and as a positive regulator of limb size during regeneration.³⁷ At a mechanistic level, we uncover adrenergic signaling as a critical mediator of both systemic activation and priming for future regeneration.

Figure 7. Main findings and implications of the study.

(A) Systemic stem cell activation following amputation promotes faster regeneration, is driven by nerves and requires adrenergic signaling. (B) Two possible models for how systemic activation may promote local limb regeneration. In both models, some limb cells (green nuclei) re-enter the cell cycle (activation) and begin to proliferate. This step occurs throughout the body. In model one, some subset of activated progenitor cells, local to the amputation site, later converts to the blastema cell state (blue cells). The subsequent migration of blastema cells to the tip of the stump creates a visible blastema (yellow). In model two, systemically-activated cells signal to nearby cells to promote functions necessary for blastema formation, which could include dedifferentiation and migration.

There are several models that may explain the regenerative role of systemically-activated cells after activation (Figure 7B). In model one, they may go on to become blastema cells, acting as progenitor cells for new tissues. In vivo lineage tracing will be necessary to fully evaluate that possibility. In model two), systemically-activated cells promote regeneration by acting as signaling hubs that influence the activity of other cells, perhaps priming them for regeneration. These models are not mutually exclusive, and possibilities for systemically-activated cells arising from traditional stem (or committed progenitor) cells or from dedifferentiation-type processes must be further investigated. While both these models focus on the role of the peripheral nervous system, potential contributions from the central nervous system, for example, the TH⁺ catecholaminergic neurons in the brain,^30,38 may also be important.

Our findings may provide potential insights into the evolution of regeneration by illustrating the possible advantage of systemic activation being a primary response to amputation. While this initial global activation event poises cells at the injury site toward a regenerative response, it simultaneously poises other tissues in the body toward accelerated regeneration should it become necessary. This global priming could have direct relevance in the wild for animals faced with intense predation—either from heterospecific or conspecific predation. Indeed, this might already have been relevant in the stem lineage of modern amphibians some 290 Ma ago, as fossil evidence not only suggests that these ancient amphibians were capable of regeneration, but also that regeneration had occurred in more than one limb in single specimens.²

Axolotls are neotenic, and, in the wild, they live their entire lives in the water; however, this neotenic life cycle is a derived trait.³⁹ For most species of salamanders, eggs are laid at very high density in a short time frame (for example, in vernal pools in the spring), resulting in high densities of larval salamanders and cannibalism.⁴⁰ Because of these circumstances, and in conjunction with our direct evidence, salamanders may need to regenerate multiple limbs in a short time frame, especially if these limbs are necessary for more mature salamanders to leave the water for their terrestrial habitats. Thus, evaluating the potential adaptive value of systemic activation to salamanders in a natural setting, while a tantalizing possibility, will require controlled experimentation, including fieldwork. The possibility that systemic activation initiates limb regeneration in wild salamanders while also preparing other appendages for future regeneration is salient.

Several other outstanding questions remain for understanding the implications of our findings. We postulate that while the initial proliferative response to amputation is global in axolotls, the sustained proliferation of activated cells and their migration out of tissue niches will be driven by local cues, potentially wound-epidermis factors (Figure 7B). Understanding how wound epidermis-derived cues influence the biology of systemically-activated cells will be important. We showed that the systemic activation and priming effects are transient. The mechanism whereby the effects are extinguished remains to be determined and defining the precise roles of epigenetics in this process requires more experimentation. There is precedence for non-immune cells, including epidermal stem cells, epigenetically encoding prior “memories” of traumatic injury and these memories sensitizing or otherwise influencing future injury responses.^41,42 Post-transcriptional and post-translational regulation of transcription factor activity in systemic activation also warrant investigation.

Systemic activation can have short-term benefits to the organism by priming other tissues for regeneration. The consequences of this system over an individual lifespan need to be evaluated. Repeated injury and chronic stimulation of this system may come at a cost. We previously demonstrated that even the axolotl’s limb regenerative abilities can be compromised by repeated amputation-regeneration cycles,⁴³ as is also the case in at least one species of newt;⁴⁴ hence, investigating how regenerative decline provoked by repeated injury interfaces with systemic activation will be interesting.

Our work suggests mammals have evolved or retained additional molecular controls that restrict natural regenerative ability. We highlight an important role for sympathetic innervation and adrenergic signaling in systemic responses to amputation, neither of which has been evaluated in mice. Future work in mice should evaluate these roles and their connection to circulating factors. These differences may reflect evolutionary differences in how regeneration responses are coordinated. Axolotls may prioritize regeneration over other physiological processes, such as growth or tissue maintenance, when necessary. This possible trade-off merits future investigation. The processes we identified may contribute to organismal resiliency in species that frequently survive intense predation, such as elephant seals that viciously fight during mating season⁴⁵ and marine mammalian species, such as dolphins, attacked by cookiecutter sharks.⁴⁶

Overall, we demonstrate that investigating the biology of systemic activation is an effective means to unravel the key pathways used to prime cells for limb regeneration. Our work predicts that activating the relevant cell types and directing them toward a regenerative response could be stimulated by harnessing adrenergic signaling in a mammalian setting, providing a major advancement toward the goal of therapeutic limb regeneration in the future. Our work offers a complementary perspective from which to consider possible effects of amputation in humans outside of the injury site. Our findings may also provide important reference points for medical applications outside of the amputee population, for example, in the context of “polytrauma,” in which a patient suffers severe wounds in several different body parts.⁴⁷

Limitations of the study

Whether epigenetic changes we observed may persist beyond the priming period remains unknown. Roles for sensory and motor peripheral innervation in systemic activation should also be explored. Potential contributions of the central nervous system are also likely to exist and should be investigated. Possible contributions of circulating epinephrine (as a function of the hypothalamus-pituitary-adrenal axis) remains to be determined. Sex-specific effects could not be evaluated, as juvenile axolotls cannot be reliably sexed morphologically, and tissue removal for molecular genotyping prior to group allocation would have confounded injury status.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Jessica Whited (jessica_whited@harvard.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• All single-cell RNA-sequencing datasets generated in this study are available at Gene Expression Omnibus (GEO), accession numbers GSE232083 (4C) and GSE232082 (2C). Bulk ATAC-seq dataset generated in this study is available at Gene Expression Omnibus (GEO), accession number GSE232080. Processed Seurat objects and reference assemblies are publicly available through the Harvard Dataverse (https://doi.org/10.7910/DVN/NW0ZN1).

• All original code has been deposited in GitHub and is publicly available through Zenodo (https://doi.org/10.5281/zenodo.17203857).

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

All animal experimentation was approved by and conducted in accordance with Harvard University’s Institutional Animal Care and Use Committee. Wild type leucistic axolotls were housed individually to ensure they remain naïve to bite injury before experimentation. Animals were housed in 2.63 g/L Instant Ocean water, 4000 μS conductivity. The animals were kept at 18°C with a 14 h light/10 h dark cycle and were fed axolotl sinking soft moist pellets (Aquatic Foods). In all experiments, and age- and size-matched (5–7 months old, 8–11 cm tail-to-snout length) clutch-mate animals were randomly assigned to each experimental group. Sex-matching of juvenile axolotls was not possible, as sex cannot be determined visually at this developmental stage. The number of replicates used in each experiment refer to number of animals unless otherwise noted.

METHOD DETAILS

Experimental Design

For all experiments, size- and age-matched clutch mates were randomly allocated to experimental groups. Sample size was estimated a priori to provide 80% power to detect a 20% difference between groups. Outcome assessments and quantitative analyses were performed blinded to group allocation. Each experimental outcome was confirmed in at least two independent experiments. For sequencing experiments, predefined inclusion criteria based on quality control were applied, and samples failing quality control were excluded from analysis.

Denervations and amputations

Animals in all experiments were both age and size matched. Animals were anesthetized in 0.1% (w/v) tricaine prior to all procedures involving amputation, denervation, or injection. After all surgical procedures, axolotls were allowed to recover overnight in 0.5% (w/v) sulfamerazine. All limb amputations were performed at the mid-stylopod and the bone was then trimmed back from the amputation plane. For forelimb denervations the brachial plexus nerves were transected according to Schotté et al. ⁶⁸ and for hindlimb denervations the sciatic nerves were transected according to Kropf et al. ⁶⁹ Briefly, incisions were made in the axial regions corresponding to the brachial plexus (forelimb) and sciatic nerves (hindlimb). Exposed nerves were then isolated with forceps and severed using fine scissors to achieve complete denervation of both forelimbs and hindlimbs. Sham operations as controls for denervations were performed by anesthetizing animals and cutting the skin open with sterile scissors, locating the nerve bundles with closed sterile forceps, but not cutting the nerves.

Drug treatments and EdU incorporation

Rapamycin was dissolved into a 25 mg/mL stock solution in ethanol. Rapamycin treatments were delivered via both water treatments at a concentration of 1 μM for 16 days with water changed daily, and through daily intraperitoneal injections at a concentration of 5 mg/kg/day for 16 days, both starting two days before amputation. All intraperitoneal injections were administered at a volume of 20 μL/g in a vehicle solution comprised of APBS with 5% Tween80, 5% PEG-400. All stock solutions were frozen at −80 °C in 1-mL aliquots. A stock solution of 6-hydroxydopamine hydrochloride (Sigma-Aldrich H438) was prepared in 0.01% ascorbic acid at a concentration of 15 mg/mL and was administered intraperitoneally for two consecutive days at a dose of 300 mg/kg. The animals were allowed to recover for 5 days before amputations. Yohimbine (Sigma-Aldrich Y3125) was dissolved into a 10 mg/mL stock solution with sterilized water. Yohimbine was administered daily at a dose of 1mg/kg for 14 days via intraperitoneal injection. Clonidine (Sigma-Aldrich C7897) was dissolved in water at 7.5 mM stock concentration and was delivered via water treatments at 30 μM. A stock solution of propranolol (Sigma-Aldrich P0884) was created by dissolving 100 mM in water. Propranolol was delivered via water treatments at 5 μM. 5-ethynyl-2′-deoxyuridine (EdU) was purchased from ThermoFisher and stock solutions were created by dissolving the powder in dimethyl sulfoxide as instructed by the manufacturer. Following anesthetization, axolotls were administered 20 μL/g of 400 μM EdU in 0.7X PBS via intraperitoneal injection 18 hours before tissue harvest.

Skeletal and tissue preparations

For immunofluorescence and HCR-FISH experiments, tissue samples were fixed in 4% paraformaldehyde post-collection in DEPC-treated PBS overnight at 4°C. Tissues were then cryopreserved in a sucrose gradient following a wash in DEPC-PBS and incubated in 30% sucrose/DEPC-PBS at 4°C overnight. Tissues were then embedded and frozen on dry ice in Tissue-Tek O.C.T. Compound (Sakura) and stored at −80°C. All blocks were sectioned at 16 μm thickness using a Leica CM 1950 cryotome and sections were stored at −80°C until use.

For wholemount RNAscope, forelimbs were fixed in 4% paraformaldehyde at 4°C overnight. Limbs were dehydrated with 50% methanol, 70% methanol, and 100% methanol. Dehydrated limbs were pre-treated with RNAscope target retrieval (ACD 322000) and protease plus (ACD 322331).

For skeletal preparations, post-collection, limbs were incubated overnight in 95% ethanol at room temperature and then incubated in 100% acetone under the same conditions. Skeletal elements were visualized with Alcian blue and Alizarin red S stains followed by a clearing of non-skeletal tissue with potassium hydroxide.

Immunofluorescence, HCR-FISH and RNAscope

Sections were rehydrated with PBS and then permeabilized with 0.5% Triton-X-100 PBS. Post-permeabilization, slides were rinsed twice with PBS and EdU visualization was performed via sulfo-cyanine 3 azide (Lumiprobe). Post-EdU visualization slides were washed twice with PBS. Staining and labeling with any additional antibodies was conducted post-EdU labeling. Primary antibodies used were rabbit anti-phospho-S6 ribosomal protein (Ser235/236, 1:200; Cell Signaling 4858), rabbit anti-tyrosine hydroxylase (1:200, Invitrogen PA5–85167). Secondary antibodies used were Cy3-goat anti-mouse (1:250, Jackson ImmunoResearch 115–165-146), Alexa Fluor 488-Goat Anti-Rabbit (1:250, Jackson ImmunoResearch, 111–545-003) and Cy3-goat anti-rabbit (1:250, Jackson ImmunoResearch 111–165-003).

Sections stained with all antibodies were blocked with 10% serum from the host species of the secondary antibody in 0.1% Triton-X-100 in PBS. Primary and secondary antibodies were diluted in in 2% BSA, 0.1% Triton-X-100 in PBS. Sections labeled with anti-pS6 and NeuN underwent an antigen retrieval step after permeabilization, in which slides were boiled in 0.1 M sodium citrate (pH 6.0) and then rinsed once in PBS, before EdU was visualized, followed by blocking and primary antibody incubation. All slides were incubated with DAPI (Sigma D9542) before cover slipping.

HCR probes were designed using the Probe Generator tool (Monaghan Lab, Northeastern University, MA, USA). HCR-FISH v3.0 was performed according to manufacturer’s instructions as described previously.⁷⁰ Briefly, hybridization chain reaction was performed on cryosectioned tissue fixed in paraformaldehyde and embedded in OCT. Sections were cleared with 4% SDS, 200mM boric acid, pH 8.5, then incubated overnight at 37 °C with initiator-labeled probe sets. Following stringent washes, hairpin amplifiers (pre-heated and snap-cooled) were added in amplification buffer and allowed to assemble overnight at room temperature to generate fluorescent signal amplification. After counterstaining with DAPI and autofluorescence quenching, sections were mounted and imaged by confocal microscopy. For RNAScope, probes were hybridized and fluorescently labeled with RNAscope Multiplex Fluorescent Detection Reagents kit (323110, ACD) following the manufacturer’s instruction. After, limbs were incubated in 1.3g/mL Histodenz (Sigma-Aldrich D2158) in PBS for up to 3 days. Samples were imaged using Zeiss lightsheet Z.1. Images were analyzed using Imaris software (Bitplane).

Cell Culture

AL-1 cells are derived from axolotl limb dermal fibroblasts^16,71. AL-1 cells were maintained as described in Lévesque et al.⁷². Cells were kept at 26°C without CO₂. They were grown in complete AL-1 media: 60% L-15 (Sigma Aldrich L5520), 5% Fetal Bovine Serum (FBS) (VWR 1500–500), 200 U/mL penicillin and 200 μg/mL streptomycin (Thermo 15–140–122), 250 ng/mL amphotericin B (Thermo 15240062), 2 mM L-Glutamine (Thermo 25–030–081), 10 μg/mL Insulin, 5.5 μg/mL transferrin, and 6.7 ng/mL sodium selenite (Thermo 41400045).

Cell Line Drug Treatments and Immunofluorescence

Cells were seeded at 100,000 cells per well in 12-well glass bottom plates. Cells were given 100 μM epinephrine diluted in water, 200nM rapamycin diluted in DMSO, vehicle control, or both 100 μM epinephrine and 200nM rapamycin. After 24 hours of treatment, cells were washed in PBS, fixed in 4% PFA for 10 minutes, rinsed in PBS, and permeabilized with 0.5% Triton-X-100 PBS. After permeabilization, cells underwent antigen retrieval where samples were boiled in 0.1 M sodium citrate (pH 6.0) and rinsed in PBS before subsequent blocking in 10% donkey serum in 0.5% Triton-X-100 PBS, overnight primary antibody incubation, secondary antibody incubation, and DAPI staining (Sigma D9542). Primary antibody used was rabbit anti-phospho-S6 ribosomal protein (Ser235/236, 1:200; Cell Signaling Technology 4858) and secondary antibody used was Alexa Fluor 488-Goat Anti-Rabbit (1:250, Jackson ImmunoResearch 111–545-003).

Imaging and quantification

Imaging was performed on a Zeiss Axio Scan.Z1 automated slide scanner, Leica M165 FC equipped with Leica DFC310 FX camera, a Nikon Eclipse Ni microscope with a DS-Ri2 camera, and a Nikon Spinning Disk Confocal (CSU-W1). Quantification of all images was conducted blindly. For imaging of homeostatic (intact), contralateral or regenerating limbs, representative sections displaying all non-epidermal tissue types were selected. For the tissue type analysis of EdU+ nuclei, tissues were classified as epidermis, “skeletal elements” (bone, cartilage), and “soft tissues” (muscle, joint, tendon, ligaments, dermis, vasculature, and nerves) based on nuclear morphology and location as previously described (Whited et al, 2013). All cell counts were conducted utilizing CellProfiler.⁷³ Measurements of blastema area were performed using Fiji.⁵⁰ For HCR signal quantifications, the wound epidermis was first masked from each image. HCR signal’s fluorescence intensity was normalized to the area of the blastema as measured by the DAPI signal.

Cell dissociation, FACS isolation, and single-cell RNA-sequencing

To isolate single cells, 12–16 homeostatic or systemically-activated limbs at 14 days post-amputation were pooled for each biological replicate and dissociated in 4 ml of dissociation buffer (100 mg/ml liberase diluted in 0.7X PBS with 0.5 U/ml DNase I) for 30 min at room temperature with gentle mixing. Suspensions were filtered through 70 μm strainers to remove cell clumps and debris, and strainers were washed with 1 ml of 0.7X PBS. Cells were spun down at 300 g at room temperature and resuspended in 1 ml of 0.7X PBS-0.04% BSA. Cells were then stained with 1 mg/mL Hoecsht, calcein AM, and propidium iodide for isolation of alive cells. Cells were further sorted into dividing (4C) and non-dividing (2C) cells by fluorescence activated cell sorting. 4C and 2C cell fractions originating from different pools of limbs were purified and were then sequenced at the Bauer Core Facility at Harvard University using an Illumina NovaSeq 6000 [https://www.illumina.com]. Paired end reads were demultiplexed with bcl2fastq2 (V2.2.1) (RRRID:SCR_015058).

ATAC-seq Library Prep and Sequencing

Ten thousand sorted 4C cells from homeostatic, systemically-activated, and regenerating limbs were resuspended in 5 μL PBS. 42.5 μL of transposition buffer (38.8 mM tris-acetate, 77.6 mM potassium acetate, 11.8 mM magnesium acetate, 0.12% NP-40, 18.8% DMF, 0.47% protease inhibitor cocktail) was added to the cells, mixed, and incubated for 10 min at room temperature. 2.5 μL of pre-loaded Tn5 (Diagenode C01070012) was added to the reaction mixture. The transposition reaction was carried out at 37°C for 30 min with shaking at 300 rpm. Following transposition, the samples were purified through DNA Clean and Concentrator-5 Kit (Zymo Research D4014). Purified library fragments were amplified with custom primers 1 and 2 using the following conditions: 72°C for 5 min; 98°C for 30 s; and thermocycling at 98°C for 10 s, 72°C for 30 s and 72°C for 1 min. The libraries were amplified for five cycles. Soon after, 5 μL aliquot of the PCR reaction was added with 10 μL of the PCR cocktail with SYBR Green (0.6X final concentration). 20 cycles of PCR was performed to determine the additional number of cycles needed for the remaining 45 μL reaction. The libraries were purified using DNA Clean and Concentrator-5 Kit (Zymo Research D4014). Purified libraries were sequenced at The Bauer Core Facility at Harvard University using an Illumina NovaSeq 6000 [https://www.illumina.com]. Paired End reads were demultiplexed with bcl2fastq2 (V2.2.1) (RRID: SCR_015058).

QUANTIFICATION AND STATISTICAL ANALYSIS

Single-cell RNA-sequencing read quantification

2C and 4C datasets were preprocessed independently and only merged after all quality control steps. To quantify reads, python v3.9.21 and the kb python package (v0.29.1) ⁷⁴ was used. Within this workflow, kallisto (v0.51.1) ⁵² was used to create a reference index and to pseudoalign reads. Bustools (v0.44.1) ⁵³ was used to quantify the output. For all steps, the latest axolotl genome assembly UKY_AmexF1_1 was used (Accession number: JBEBLI000000000.1, RefSeq assembly: GCF_040938575.1) (https://ambystoma.uky.edu/genome-resources). Matrix files were then read into R Studio (v 4.3.3) and a Seurat object (v5.0.3) ⁵⁴ was created. Empty droplets were filtered out using the emptyDrops command within DropletUtils (v1.22.0) ⁵⁵, where an FDR >0.001 was chosen as a cutoff value. Further, cells with <500 UMIs were removed, to only include cells with adequate sequencing depth. Data was clustered preliminarily in Seurat as input to SoupX (1.6.2) ⁵⁶ to correct for ambient RNA contamination. SoupAdjusted counts were used to filter out doublets via the scDblFinder package ⁵⁷ (v1.16.0). Finally, dying cells (>5% mitochondrial gene counts) were removed.

2C and 4C datasets were independently normalized on their respective SoupAdjusted layer using the SCTransform command ⁷⁵, and potential mitochondrial mapping confounding variables were accounted for via the vars.to.regress parameter. 2C and 4C datasets were finally merged and integrated using Harmony ⁵⁸ (v1.2.0). Cells were clustered in Seurat with 15 dimensions and at a resolution of 0.15 and visualized with the UMAP reduction method. Marker genes were determined via thze FindAllMarkers command, where only genes expressed in >20% of cells in a given cluster, with an associated log2 fold-change >0.25 were considered. These marker genes were used to annotate clusters manually.

To determine co-expression of Adra2a with blastema marker genes, cells were categorized into four groups. If cells expressed >0 counts of Adra2a, they were flagged as Adra2a+, while all other cells were classified as Adra2a-. Cells were further divided into homeostatic and activated, dependent on their experimental condition. Next, the Seurat command FindAllMarkers was deployed to test whether common blastema marker genes ⁷⁶ were differentially expressed, with a significance-threshold of 0.01. Data was further visualized via the DotPlot command within Seurat.

ATAC-seq read processing

We aligned sequencing reads to the reference genome (AMEX60DD) using Bowtie2 (v2.3.2)⁵⁹ with parameters optimized for local alignment (--no-mixed), high sensitivity (-X 2000), gapped alignment (--dovetail), and exclusion of discordant read pairs (--no-discordant). Alignments were sorted and indexed throughout the analysis using Samtools (v1.10)⁶⁰ for efficient retrieval. Mitochondrial reads were filtered out using grep to focus on nuclear DNA.Low-quality alignments (MAPQ score < 30) were removed using Samtools. PCR and optical duplicates were identified and removed with Picard MarkDuplicates⁶¹ [https://broadinstitute.github.io/picard/] assuming sorted alignments (ASSUME_SORTED=true) and removing all identified duplicates (REMOVE_DUPLICATES=true). Generation of bigwig Coverage Tracks was performed with deepTools bamCoverage⁷⁷. We specified --outFileFormat=bigwig for bigWig output, - normalizeUsing CPM for counts per million normalization, and --binSize 20 for a bin size of 20 base pairs.

ATAC-seq peak calling

Open chromatin regions were identified using MACS2 (v2.2.9.1)⁶² (https://github.com/macs3-project/MACS). We employed two peak calling strategies: consensus peaks, generated by analyzing all replicates/samples together, and condition-specific peaks.

Annotating and visualizing the ATAC data

We employed ChIPseeker (v1.32.1)⁶³ to annotate the peaks. Subread featurecount (v1.32.1)⁷⁸ was used to create a gene by peak count matrix. Differential accessibility was observed using DESeq2.⁶⁴ To visualize normalized counts and sample clustering, we used Pheatmap (v1.9.12) (10.32614/CRAN.package.pheatmap) ⁶⁵ and pcaExplorer (v2.22.0)⁶⁶, respectively. Motif discovery and enrichment analysis were performed using HOMER (v4.11).⁶⁷ Briefly, motifs from HOMER analysis were read into MATLAB and significant motifs identified in each replicate for each condition. Replicates with number of enriched motifs significantly lower than in other replicates for the same condition were discarded in further analyses. 4C single-cell gene expression linked to discovered motifs was visualized with Seurat’s DotPlot function.⁷⁹ We preprocessed the 4C data using Seurat’s AggregateExpression function with log normalization and scaling. Heatmaps of the aggregated 4C single cell counts were generated with pheatmap (10.32614/CRAN.package.pheatmap)⁶⁵. Finally, pathway enrichment analysis of the transcription factors was performed using Panther⁸⁰ and visualized with ggplot2.⁸¹

Supplementary Material

1

Figure S1. Denervation surgery (Related to Figure 2). Cartoon depicting nerves innervating the axolotl limb in the naïve state (left) and the sham-operation procedure (right).

2

Figure S2. Activation of α-adrenergic signaling is not sufficient for systemic activation (Related to Figure 3). (A) EdU stained intact limbs harvested from animals treated with Adra2a agonist clonidine or vehicle solution for 9 days, quantified in (B) (vehicle n=14, clonidine n=12). (C-D) Animals treated with Adra2a agonist clonidine or vehicle solution were unilaterally denervated followed by proximal amputations. (C) EdU stained limbs contralateral to denervated and amputated limbs, harvested at 14 dpa, quantified in (D) (vehicle n=19, clonidine n=17). (E-F) Animals treated with Adra2a agonist clonidine or vehicle solution were unilaterally denervated on one side, followed by unilateral proximal amputations of the innervated limbs. (E) EdU staining of limbs, denervated and contralateral to unilateral amputations, harvested at 14 dpa, quantified in (F) (vehicle n=17, clonidine n=15). Statistical significance was determined using unpaired two-tailed Welch’s t-test in (D) and (F). Scale bars, 500 μm.

3

Figure S3. Marker genes used to identify clusters in single-cell RNA sequencing of 4C and 2C cells (Related to Figure 6). Dot plots showing the relative frequency of expression (dot size) and expression level (color) of marker genes across all clusters.

4

Figure S4. Expression of adrenergic receptors across cell types (Related to Figure 6). (A-B) Dot plots showing the relative frequency of expression (dot size) and expression level (color) for alpha- (A) and beta- (B) adrenergic receptors across all cell clusters.

5

Figure S5. Expression of Kazald2 and Snai1 in the wholemount homeostatic limb (Related to Figure 6). Representative images of wholemount limbs showing Kazald2 and Snai1 expression in the skin (top panel), muscle (middle panel) and bone (bottom panel).

6

Supplementary Movie 1. Visualization of whole mount bone of a homeostatic axolotl limb showing Kazald2 (red) and Snai1 (green) expression (Related to Figure 6).

7

Table S1. List of transcripts expressed by each cell cluster in homeostatic and systemically-activated limbs (Related to Figure 6).

8

Table S2. Differentially expressed transcripts expressed by each cell cluster in homeostatic and systemically-activated limbs (Related to Figure 6).

9

Table S3. List of accessible transcription factor binding motifs enriched across 4C cells of homeostatic, activated, and regenerating limbs. (Related to Figure 6).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-phospho-S6 ribosomal protein (IF 1:200)	Cell Signaling Technology	Cat#: 4858
Rabbit anti-tyrosine hydroxylase (IF 1:200)	Invitrogen	Cat#: PA5–85167
Cy3-goat anti-mouse (IF 1:250)	Jackson ImmunoResearch Labs	Cat#: 115–165-146
Alexa Fluor 488-Goat Anti-Rabbit (IF 1:250)	Jackson ImmunoResearch Labs	Cat#: 111–545-003
Cy3-goat anti-rabbit (IF 1:250)	Jackson ImmunoResearch Labs	Cat#: 111–165-003
Chemicals, peptides, and recombinant proteins
6-hydroxydopamine	Sigma-Aldrich	H4381
Yohimbine	Sigma-Aldrich	Y3125
Clonidine	Sigma-Aldrich	C7897
Propranolol	Sigma-Aldrich	P0884
Rapamycin	LC Laboratories	R-5000
Adrenaline	Santa Cruz Biotechnology	sc-252780
5-ethynyl-2′-deoxyuridine (EdU)	Thermo Fisher Scientific	E10187
DAPI	Sigma-Aldrich	10236276001
BSA	Sigma-Aldrich	A7906
Liberase™ TM Research Grade	Sigma-Aldrich	5401119001
DNase I	Invitrogen	18047019
Hoechst 33342	Thermo Scientific	62249
Calcein AM	Invitrogen	C3099
Propidium iodide	Invitrogen	P3566
Tn5	Diagenode	C01070012
Deposited data
4C Single-cell RNA-seq datasets	This manuscript	GEO: GSE232083
2C Single-cell RNA-seq datasets	This manuscript	GEO: GSE232082
Bulk ATAC-sequencing datasets	This manuscript	GEO: GSE232080
Code for sequencing data analysis	This manuscript	https://doi.org/10.5281/zenodo.17203857
Processed Seurat objects and reference assemblies	This manuscript	https://doi.org/10.7910/DVN/NW0ZN1
Software and algorithms
Zen version 3.3	ZEN Digital Imaging for Light Microscopy (RRID:SCR_013672)	https://www.zeiss.com/microscopy/en/products/software/zeiss-zen.html
GraphPad Prism v9.2.0	Dotmatics	https://www.graphpad.com
CellProfiler v4.3.1	Carpenter et al.49	https://cellprofiler.org
Fiji 2.0	Schindelin et al 50	https://fiji.sc
HCR probe generator	Monaghan Lab (MA, USA)	http://ec2-44-211-232-78.compute-1.amazonaws.com
Python v3.9.21	Python Software Foundation	https://www.python.org/
Python kb package v0.29.1	Melsted et al. 51	https://pypi.org/project/kb-python/
kallisto v0.51.1	Bray et al.52	https://pachterlab.github.io/kallisto/
Bustools v0.44.1	Melsted et al.53	https://github.com/BUStools/bustools/releases
R v4.2.1	CRAN	https://www.r-project.org/
R Studio v4.3.3	Posit PBC	https://posit.co/
Seruat v5.0.3	Hao et al.54	https://satijalab.org/seurat/
DropletUtils v1.22.0	Lun et al.55	https://bioconductor.org/packages/release/bioc/html/DropletUtils.html
SoupX v1.6.2	Young et al.56	https://cran.r-project.org/web/packages/SoupX/index.html
scDblFinder v1.16.0	Germain et al.57	https://bioconductor.org/packages/release/bioc/html/scDblFinder.html
Harmony v1.2.0	Korsunsky et al.58	https://github.com/immunogenomics/harmony
Bowtie2 v2.3.2	Langmead et al.59	https://bowtie-bio.sourceforge.net/bowtie2/index.shtml
Samtools v1.10	Li et al.60	https://github.com/samtools/samtools/releases/
Picard MarkDuplicates	Picard Toolkit61	https://broadinstitute.github.io/picard/
MACS2 v2.2.9.1	Zhang et al.62	https://github.com/macs3project/MACS
ChIPseeker v1.32.1	Wang et al.63	https://www.bioconductor.org/packages/release/bioc/html/ChIPseeker.html
DESeq2	Love et al.64	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
pheatmap v1.9.12	Kolde65	https://cran.r-project.org/web/packages/pheatmap/index.html
pcaExplorer v2.22.0	Marini et al.66	https://www.bioconductor.org/packages/release/bioc/html/pcaExplorer.html
HOMER v4.11	Heinz et al.67	http://homer.ucsd.edu/homer/

HIGHLIGHTS

• Systemic stem cell activation following amputation promotes faster regeneration

• Sympathetic nerves are required for systemic cell cycle re-entry and limb regeneration

• Adrenergic - mTOR axis drives systemic response to amputation

• Systemically activated cell types are shared with proliferating cell types in homeostasis