Regulation of cell cycle progression is essential for cell proliferation during regeneration following injury. After appendage amputation, the axolotl (Ambystoma mexicanum) regenerates missing structures through an accumulation of proliferating cells known as the blastema. To study cell division during blastema growth, we generated a transgenic line of axolotls that ubiquitously expresses a bicistronic version of the fluorescent ubiquitination-based cell-cycle indicator (FUCCI). We demonstrate near-ubiquitous FUCCI expression in developing and adult tissues, and validate these expression patterns with DNA synthesis and mitosis phase markers. We demonstrate the utility of FUCCI for live and whole-mount imaging, showing the predominantly local contribution of cells during limb and tail regeneration. We also show that spinal cord amputation results in increased proliferation at least 5 mm from the site of injury. Finally, we use multimodal staining to provide cell type information for cycling cells by combining fluorescence in situ hybridization, EdU click-chemistry and immunohistochemistry on a single FUCCI tissue section. This new line of animals will be useful for studying cell cycle dynamics using in situ endpoint assays and in vivo imaging in developing and regenerating animals.
Vertebrate tissue regeneration inherently requires cell proliferation either through endogenous stem cell proliferation or re-entry of differentiated cells into the cell cycle. One of the most striking examples of vertebrate regeneration is epimorphic replacement of the amputated salamander appendage. Appendage regeneration requires the generation of a highly proliferative mass of cells called the blastema. The formation of the blastema is dependent on an intact nerve supply and a specialized layer of epithelium known as the apical epithelial cap (AEC) (McCusker et al., 2015a). The AEC likely has multiple functions, including directing outgrowth, maintaining proliferation and secreting factors that allow for remodeling of the underlying extracellular matrix (Stocum, 2017; Tsai et al., 2020). Although the blastema consists of numerous cell types, most cells originate from mesenchymal cell populations located near the amputation plane (Butler, 1933; Currie et al., 2016). Understanding the mechanisms that initiate and sustain proliferation of blastema cells is a fundamental problem that requires modern molecular tools to track and characterize blastema cell behavior (Stocum, 2017; Tanaka, 2016). Although recent developments in transgenesis and tissue grafting techniques have allowed the observation of blastema cells in vivo (Currie et al., 2016; Khattak et al., 2013; Kragl and Tanaka, 2009; Sandoval-Guzmán et al., 2014), further development of transgenic lines are needed to enable imaging of the regeneration process.
In 2008, Sakaue-Sawano and colleagues developed the fluorescent, ubiquitination-based cell-cycle indicator (FUCCI) system to study cell cycle progression in human cell lines and mice (Sakaue-Sawano et al., 2008). Since then, several variations have been made to the FUCCI construct, and it has been used to generate transgenic plants (Yin et al., 2014) and animals (Abe et al., 2013; Sugiyama et al., 2009; Zielke et al., 2014). The FUCCI system is based upon the inverse oscillation of Geminin and Cdt1 proteins that occur naturally during the cell cycle (Nishitani et al., 2004). The FUCCI construct includes a constitutively active promoter that drives expression of a fluorescent protein fused to the Cdt1 protein degron, which has high levels in G1 phase due to ubiquitin-mediated proteolysis during S, G2 and early mitotic phases (S/G2/M). Conversely, the Geminin protein degron is fused to a different fluorescent molecule that is degraded during late M and G1, leading to high fluorescent protein levels in S/G2/early M (see Zielke and Edgar, 2015 for a review). Fusion of these two expression cassettes into a single bicistronic transgene allows visualization of cells while they progress through the cell cycle (Bouldin and Kimelman, 2014). In this study, we use a ubiquitously expressed bicistronic FUCCI construct to generate a transgenic line of axolotl salamanders to study cell cycle dynamics during development, limb regeneration and tail regeneration.
Generation and characterization of developing FUCCI axolotls
We chose to design a bicistronic version of the original FUCCI construct because it should theoretically lead to equimolar levels of probes and only require the generation of a single transgenic animal line (Rajan et al., 2018). The construct includes a CAG promoter that drives expression of monomeric Azami-Green fused to the zebrafish geminin degron (mAG-zGem) followed by a viral 2A self-cleaving peptide and mCherry fused to the zebrafish Cdt1 degron (mCherry-zCdt1), which was cloned into the pISce-Dest backbone using Gateway cloning (Fig. 1A) (Kwan et al., 2007). F0 animals were generated using standard axolotl injection conditions with I-SceI Meganuclease (Khattak et al., 2009), generating F0 animals with mosaic FUCCI expression. A single female was selected due to strong ubiquitous FUCCI expression (Fig. 1B) to mate with a d/d white male, which generated clutches consisting of 82% and 92.7% transmission rates, suggesting the FUCCI construct integrated multiple times in the founder animal. To confirm this, we crossed F1 FUCCI animals to a F1 FUCCI sibling or a d/d white animal. In each cross, we observed a non-mendelian distribution of offspring (Table S1), confirming the presence of multiple integrations in the F1 generation. F2 larvae exhibit similar mAG and mCherry expression patterns to the F1 generation throughout embryonic development (data not shown). All subsequent experiments in the following sections were conducted on FUCCI animals from the F1 generation.
Examining live transgenic embryos, we first detected FUCCI protein expression at neurulation with increasing expression throughout development (Fig. 1C,D). Expression was variable between siblings, possibly due to varying levels of transcriptional activation or the presence of multiple integrations of the FUCCI construct into the genome (Fig. 1D-F). Gross observation of transgenic larvae clearly showed distinct non-proliferative G1 populations, including the somites (Fig. 1D), the tail myotomes (Fig. 1G), lateral line neuromasts (Fig. 1G) and highly proliferative S/G2/M populations, such as the limb bud and larval heart (Fig. 1H). To determine the adult expression pattern of F1 FUCCI offspring, tissues sections were analyzed from the brain, eye, heart, liver, spleen and gut. mAG and mCherry expression were observed in every tissue type with little overlap between probes except for differentiated muscle fibers (Fig. 1I-N, Fig. S1).
FUCCI probes overlap with S- and M-phase markers, and correlate with DNA content
To determine the overall expression level of the FUCCI construct in F1 animals, we quantified 2547 cells from FUCCI spinal cords pulsed with ethynyldeoxyuridine (EdU) for 3 h (n=9). In total, 22.34% of the cells were mAG+/mCherry−, 71.22% were mAG−/mCherry+, 2.75% were mAG+/mCherry+ and 3.69% mAG−/mCherry− (Fig. S2A). Interestingly, we observed few mAG+/mCherry+ cells. This observation is corroborated in similar FUCCI axolotl lines (Costa et al., 2021), and is likely because the cells are in a regenerative environment where they are receiving mitogens instructing them to enter the cell cycle instead of resting at the G1/S boundary (Takahashi et al., 2021). We next determined whether mAG+ cells were specific to the S phase by performing click-it based EdU detection of DNA synthesis on the same regenerating spinal cord tissues. Fig. 2A illustrates the expected pattern of labeling for mAG, mCherry, EdU and phosphorylated serine 10 histone H3 (pHH3). Of the 532 EdU+ cells (20.9% of total cells), 88.93% were mAG+/mCherry−, 3.00% were mAG+/mCherry+, 1.88% were mAG−/mCherry+ and 6.19% were mAG−/mCherry− (Fig. 2B, Fig. S2D). Conversely, 76.68% of mAG+ cells were EdU+ and 23.32% of mAG+ cells were EdU− (Fig. S2B,D), suggesting that the majority of mAG+ cells were in S phase rather than G2/M phases. This suggests that the S phase is longer than the combined G2/M phases by approximately threefold, which is supported by previous studies (McCullough and Tassava, 1976). To study mitosis in a highly proliferative tissue, we performed immunohistochemistry with pHH3 in 10 dpa regenerating limb blastemas (n=3) (Fig. S2E-H″″). We found that 88.89% of pHH3+ cells were mAG+/mCherry−, 4.44% were mAG+/mCherry+, 0% were mAG−/mCherry+ and 6.67% were mAG−/mCherry− (Fig. 2C). Furthermore, we found that 3.91% of mAG+ cells were pHH3+ whereas 96.09% were pHH3− (Fig. S2C), suggesting that the vast majority of mAG+ limb blastema cells are in S/G2 phases. As expected, some pHH3+ had neither probe signal, as Geminin is known to degrade in the late stages of M phase (McGarry and Kirschner, 1998). Overall, the high correspondence between EdU and pHH3 with mAG expression shows that mAG-zGem effectively marks cells in both S and M phases.
We next identified cells at each stage of the cell cycle according to DAPI and EdU staining. Each stage of the cell cycle was observed in proliferating tissues and had predictable genomic structure, EdU incorporation (Leonhardt et al., 2000), pHH3 staining and FUCCI reporter expression (Fig. 2D-M). Importantly, we observe mAG+/mCherry+/EdU+ cells (Fig. 2H), indicating the presence of cells in very early S phase. These cells have not yet fully degraded mCherry but have started translating mAG while incorporating EdU in early S phase. These cells have not yet fully degraded mCherry-zCdt1 but have switched from degrading mAG-zGem to mCherry-zCdt1. Collectively, these results demonstrate that our FUCCI construct correctly and reproducibly labels specific stages of the cell cycle.
To correlate DNA content with FUCCI probes, we dissociated FUCCI limb blastemas and stained the single-cell suspension with Hoechst DNA stain. These cells were then analyzed with FACS based on FUCCI probe fluorescence and Hoechst intensity (Fig. S3A-B″). Imaging both unsorted and sorted cells within the mAG+/mCherry−, mAG−/mCherry+ and mAG+/mCherry+ gates showed pure populations of the respective fluorescent signal (Fig. S3C-F‴). We found that 84% of the mAG−/mCherry− population was within the 2N gate, while 65.8% of the mAG+/mCherry+ population was within the 4N gate, validating the 2N and 4N gates (Fig. S3B″). We found that 69% of the mAG−/mCherry+ (G1) population was within the 2N gate, and 30.1% was within the 4N gate (Fig. S3B″). These percentages shifted in the mAG+/mCherry− (S/G2/M) population, where we observed that 59.5% and 40.2% of cells fell within the 2N and 4N gates, respectively (Fig. S3B″). Although the shift in DNA content fits the expected trend for each population, we had expected there to be a larger percentage of both the G1 population within the 2N gate and the S/G2/M population to be within the 4N gate. We suspect this is due to heterogeneity within the blastema population for two reasons. First, bright-field images of the pre-sorted single celled suspension reveal cells of different sizes and granularity, all with varying degrees of mCherry and mAG fluorescence (Fig. S3C-C‴). Second, the histogram of Hoechst staining in the entire blastema population does not have two well-defined 2N and 4N peaks (Fig. S3B). Together, these data indicate that the different cells within the blastema mesenchyme may vary in their Hoechst fluorescent signal in both the 2N and 4N states. Future studies evaluating the FUCCI signal within specific cell types that contribute to the blastema may resolve this heterogeneity issue.
In vivo imaging of FUCCI expression
To determine the feasibility of in vivo imaging of FUCCI tissue, we imaged cycling epithelial cells on the surface of an anesthetized stage 32 larva mounted in 0.3% agarose (Movie 1). Fig. 3A-E depicts a single magnified mAG+ cell from Movie 1 completing the process of mitosis in under 30 min to produce two daughter cells with fading mAG intensity (Fig. 3A-E, Movie 2). During this process, we observe the formation of the mitotic spindle in prophase and cytokinesis after chromosome separation. We did not observe any cell transition from mAG+/mCherry− to mAG−/mCherry+, or vice versa. This is likely due to the shortening of the G1 phase during embryonic development (Siefert et al., 2015), which would prevent the accumulation of mCherry protein to provide a detectable signal while transitioning from M phase or before transitioning to S phase.
To visualize blastema formation in vivo, we performed live imaging of a regenerating tail in a stage 36 FUCCI animal (Fig. 3F-I). After amputation, the anesthetized larva was immediately embedded in 0.3% agarose and imaged every 30 min over 60 h (Movie 3). During this 60 h imaging experiment, we observed early wound healing (Fig. 3G), blastema formation (Fig. 3H) and myomeric muscle development (Fig. 3I). By 8 h post-amputation (hpa), the tail stump was completely covered by a thin layer of both mAG+ cells and mCherry+ cells (Fig. 3G). Shortly afterwards, an early blastema was observed in the posterior tail tip by 17.5 hpa (Fig. 3H) that mostly comprised mAG+ cells. At this time point, early myomeric muscle formation was observed along the anteroposterior (AP) axis of the tail, characterized by regularly spaced bar-shaped groups of mCherry+ cells (Fig. 3H,I). By 37.5 hpa, mAG+ cells at the amputation plane started accumulating at the base of the blastema (Fig. 3I). After ∼40 h of imaging, cells in the blastema seemed to be dying. However, subsequent time points showed continued maturation of the tail myomere muscle and a general shift from mostly mAG+ cells in the tail to mCherry+ cells.
To track migrating and dividing mAG+ cells in the regenerating tail, we used TrackMate (Tinevez et al., 2017). With this, we tracked the position of mAG+ cells 10 h before and after each frame (Fig. 3J-M, Movie 4). This allowed us to visualize the path cells took to contribute to the regenerated tail. Interestingly, during the early blastema formation phase, we observed a general trend for dorsal tail cells to migrate dorsally, ventral cells to migrate ventrally, and cells in the midline to migrate in the direction of the blastema (Fig. 3J,K). At later time points in the movie, we observed that the intense mAG+ cells adjacent to the amputation plane were migrating into the tail blastema (Fig. 3M).
We quantified changes in fluorescence of mAG by dividing the regenerating tail into rectangles with a 30 µm width anterior and posterior to the amputation plane (Fig. S4). The raw integrated density for each channel was measured for each box and normalized to the total tail area within each rectangle, providing a measure of intensity per area. These results showed an increase in the intensity of mAG fluorescence in rectangles starting 30 µm anterior to the amputation plane and continuing into the regenerating tail tip, indicating that proliferation is highest in the blastema and in cells 30 µm anterior to the amputation plane (Fig. 3N). Anterior to the amputation plane, the mAG intensity was higher at the earliest time points and steadily decreased after 60 h of imaging (Fig. 3N). The opposite trend was observed for mCherry fluorescence, where the intensity increased after 60 h of imaging (Fig. 3O). These results indicate an increase in the total number of cells in G1 phase, which may represent an accumulation of cells in the resting state after rapid proliferation during early embryonic development.
Multimodal imaging provides cell type identity to FUCCI tissues
The use of two fluorescent proteins in FUCCI sensors limits the acquisition of cell type information, precluding robust cell characterization using imaging modalities including immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH) in conjunction with FUCCI probes. To overcome this limitation, we observed that mAG and mCherry can be sufficiently photobleached after imaging to allow for multimodal imaging (Fig. 4A). We first performed version 3 hybridization chain reaction FISH (V3.HCR-FISH) (Choi et al., 2018) for Shh using Alexa-fluor 647 on an EdU pulsed, homeostatic FUCCI spinal cord (Fig. 4B, Fig. S5A-A″″). We next photobleached the endogenous FUCCI signal and removed the Shh probes with 80% formamide (Fig. 4C, Fig. S5B-B″″). Then a subsequent round of V3.HCR-FISH was performed for Pax7 and B3Tub (Fig. 4D, Fig. S5C-C″″). Imaging and subsequent removal of these probes were followed by EdU labeling (Fig. 4E, Fig. S5D-D″″). EdU signal was then removed with DNase, and IHC was performed for B3TUB (Fig. 4F, Fig. S5E-E″″). The images from each round were aligned to the DAPI image from the first round, allowing imaging of four modalities (transgenic reporter, FISH, click-chemistry and IHC) in the same tissue section (Fig. 4G-J). This analysis shows that cell type identification can be performed along with the study of cell division in FUCCI tissue.
Regenerating FUCCI limbs reveal distinct regions of proliferative and non-proliferative zones in the limb blastema
To visualize cell cycling following limb amputation, we imaged uninjured limbs (Fig. 5A,A′) and regenerating limbs from five animals at 1, 3, 5, 7, 10, and 14 dpa (Fig. 5B-G′). To quantify the location of proliferation following amputation, we calculated the average distance between the amputation plane and the distal mCherry+ muscle boundary line. We found that mAG+ cells were abundant proximal to the amputation plane as early as 1 dpa, and that these mAG+ cells were located on average 243.85 µm proximal to the amputation plane (Fig. S6A). This distance from the amputation plane was significantly larger than the same measurement at 5, 7, and 14 dpa (one-way ANOVA with a Tukey-Kramer multiple comparisons, P<0.05). These findings correspond well with previous studies, where it was determined that cells located up to 500 µm proximal to the amputation plane participate in regeneration (Currie et al., 2016). As the limb regenerates, we observed fewer mAG+ cells proximal to the amputation as more mAG+ cells accumulated within the blastema (Fig. S6A). This observation suggests that the supply of proliferating cells becomes less dependent on cells proximal to the amputation plane at later time points during regeneration.
To determine the cell type of the mAG+ and mCherry+ cells in regenerating limbs, we visualized FUCCI probe expression in whole-mount with light-sheet fluorescence microscopy (Fig. 5H-N, Fig. S6B-J, Movies 5 and 6). Based on cell morphology, we observed that the majority of uninjured tissue, including fibroblasts, epithelial cells and chondrocytes were mCherry+. Most muscle cells observed were mAG+/mCherry+, which is consistent with a similar G1/S arrest in FUCCI mouse cardiomyocytes (Alvarez et al., 2019). Very few mAG+ cells were observed in uninjured tissue (Fig. S6C, Movie 5). At 1 dpa, we observed several mAG+ cell types, including the wound epithelium, perichondrium and some fibroblasts of the mesenchyme (Fig. S6D). These cell types appear to remain mAG+ until blastema formation at 7 dpa, where fewer chondrocytes, perichondrial cells and epithelial cells are mAG+ (Fig. S6E-G). From 7 to 14 dpa, most of the mAG+ cells are located within the mesenchyme, further showing that cells proximal to the amputation plane less frequently proliferate at later time points during limb regeneration (Fig. S6H,I). By 21 dpa, most of the mAG+ cells are chondrocytes in the regenerating carpals and digits (Fig. S6B,J, Movie 6). At 10 and 14 dpa, we observed a small population of cells at the distal-most tip of the blastema that were mAG−/mCherry+ (Fig. 5F′-G′). We sectioned EdU-pulsed 10 dpa blastemas for histological analysis (n=3) and found that this mAG−/mCherry+ population comprised the distal-most epithelial cells of the AEC (Fig. S7A-B″″). In one sample, a small number of these cells was observed in the distal-most region of the blastema mesenchyme (Fig. S7A). We then sectioned 14 dpa blastemas (n=3) and observed the presence of these distal-most, mesenchymal mAG−/mCherry+ cells in all three samples (Fig. S7C-E‴), suggesting that this population of cells is more abundant at later stages of regeneration. The observation that the highest proliferation levels are located in the middle-proximodistal region of the blastema has been observed by others in salamanders (Farkas et al., 2016; McCusker et al., 2015b) and in regenerating zebrafish fins (Hirose et al., 2014).
For quantification of blastema cells, we performed flow cytometry analysis on 10 dpa FUCCI blastemas (n=10) (Fig. S8A-D). In total, 5682 cells were analyzed from the total 10,000 events. Of these cells, 25.2% mAG+/mCherry−, 53.9% were mAG−/mCherry+, 1.1% were mAG+/mCherry+ and 16.3% were mAG−/mCherry−. Although these results do not exactly correspond with our tissue section quantification (n=3), where we found that 44.68% were mAG+/mCherry−, 14.96% were mAG−/mCherry+, 14.06% were mAG+/mCherry+ and 26.3% were mAG−/mCherry−, it is not surprising considering that some mCherry+ muscle cells immediately proximal to the blastema were collected for dissociation (Fig. S8E). Additionally, the amount of autofluorescent blood, as well as the cessation of proliferation that likely occurs as the cells are dissociated before FACS analysis, may contribute to the discrepancy.
Blastema cells arrest in G1 phase in the absence of an intact nerve supply
Transection of the nerve supply to the limb inhibits limb regeneration (Farkas and Monaghan, 2017; Todd, 1823). In the absence of nerves, cell proliferation, DNA synthesis rates and the relative number of mitoses in regenerating limbs are all lower than in innervated limbs (Duerr et al., 2020; Kumar et al., 2007; Loyd and Tassava, 1980; Mescher and Tassava, 1975; Monaghan et al., 2009; Tassava et al., 1974). However, our knowledge of how nerves regulate the cell cycle is incomplete. To better understand the impact of denervation on cell cycle regulation in the blastema, we performed FACS analysis on innervated and denervated FUCCI blastemas (Fig. 6). Forelimbs were bilaterally amputated and, at 17 dpa (early/mid bud stage), right limbs were denervated at the brachial plexus while left limbs remained innervated (n=11 blastemas per condition). To ensure enough time for a cellular response to denervation, blastema cells were dissociated 48 h after denervation (Fig. 6A-B‴) and the number of mAG+/mCherry−, mAG−/mCherry+, mAG+/mCherry+ and mAG−/mCherry− cells from innervated and denervated blastemas (wound epithelium removed) were quantified with FACS (Fig. 6C,D). Our data indicate that denervation decreases the number of mAG+/mCherry− blastema cells from 17% in innervated to 15% in denervated, with an increase in the number of mAG−/mCherry+ cells from 64% in innervated to 66.2% in denervated (Fig. 6E,F). Furthermore, the number of mAG+/mCherry+ cells decreases from 16.5% in innervated limbs to 14.7% in denervated, and the number of mAG−/mCherry− cells increases from 2.4% in innervated to 4.2% in denervated (Fig. 6E,F). Together, these results indicate that, in the absence of nerves, some blastema cells arrest in G1 phase (decrease in the number of mAG+/mCherry− and mAG+/mCherry+ cells, increase in the number of mAG−/mCherry+ cells) or exit the cell cycle (increase in the number of mAG−/mCherry− cells).
Spinal cord amputation induces a proliferation response 5 mm from the injury
To determine the location of proliferating cells along the AP axis of the regenerating spinal cord, we collected 14 dpa EdU pulsed FUCCI spinal cords (n=4) and obtained tissue sections at various locations along the AP axis in relation to the most posterior tip of the regenerated cartilaginous rod (Fig. 7A, Table S2). The amputation plane is located between the 250 µm anterior and 500 µm anterior sections, as the notochord was identified 500 µm anterior to the cartilaginous rod but not at the 250 µm anterior section. For comparison, we sectioned spinal cords from non-regenerating, homeostatic FUCCI animals (n=5) (Table S2). Quantification included cells within the boundary of the meninges that surrounds the spinal cord but excluded the meningeal cells themselves (Fig. 7B). mAG+/mCherry− cells were consistently around 40% of total cells posteriorly and progressively declined anteriorly (Fig. 7C-L), which may indicate that cell proliferation is most abundant at or posterior to the cartilaginous rod tip. This is accompanied by an increase in the number of mAG−/mCherry+ cells anteriorly along the regenerating AP axis (Fig. 7C-L), suggesting a shortened G1 phase in regenerating cells that is supported by previous studies (Rodrigo Albors et al., 2015). Furthermore, we observed a significant increase in the number of mAG+/mCherry− cells located 5000 µm from the regenerated cartilaginous rod compared with uninjured spinal cords (two-tailed Student's t-test assuming unequal variances, P=0.0043), suggesting that spinal cord injury induces an increase in cell cycling beyond 500 µm anterior to the amputation plane (Fig. 7L, Table S2). Our results also indicate that the relative abundance of cells in S or G2/M, as indicated by mAG+/EdU+ or mAG+/EdU−, respectively, is unchanged across the AP axis (Fig. 7M, Table S2), suggesting that the ratio of S:G2 does not significantly change across the regenerating AP axis. However, a significant difference is detected between the total number of mAG+/EdU+ cells detected in sections 5000 µm from the cartilaginous rod tip and uninjured spinal cords (two-tailed Student's t-test assuming unequal variances, P=0.001) (Fig. 7M). A statistically significant difference between the number of mAG+/EdU− cells at these regions was not detected (two-tailed Student's t-test assuming unequal variances, P=0.143) (Fig. 7M). Taken together, our results indicate that spinal cord injury induces an increase in the number of cycling cells along the AP axis nearly 5 mm from the injury compared with uninjured controls.
The use of FUCCI sensors in the axolotl enables the study of cell cycle dynamics during tissue regeneration that were previously hindered without an in vivo marker of cell state. By using the FUCCI system, we validated findings from previous studies, including the mostly local recruitment of cells required for limb (Butler, 1933) and tail regeneration (McHedlishvili et al., 2007). We also uncovered new findings regarding the impact of the nerve on the cell cycle during limb regeneration, showing that blastema cells arrest in G1 phase after severing the nerve supply. This could be explained by the absence of mitogenic factors supplied by the nerve, such as NRG1 (Farkas et al., 2016), transferrin (Mescher and Munaim, 1984) and FGFs/BMPs (Makanae et al., 2014). We also show the presence of proliferative ependymal cells in the spinal cord almost 5 mm from the site of injury. Several explanations exist for the proliferative response in the spinal cord, including a systemic proliferation response after injury (Johnson et al., 2018), an organ-wide proliferation response as seen in the regenerating lung (Jensen et al., 2021) or proliferation induced by neuronal death (Hedlund et al., 2016). These results demonstrate the power of FUCCI sensors for the study of cell cycle dynamics during tissue regeneration. Recently, others have developed a similar FUCCI system in the axolotl to explore cell cycle dynamics in the regenerating spinal cord (Costa et al., 2021). These FUCCI animal lines will provide an important tool to the community to better understand how the cell cycle is regulated in the axolotl during tissue regeneration. This knowledge is crucial for informing regenerative medicine to induce a more-robust proliferative response in mammals after spinal cord injury or limb amputation.
Many fundamental questions remain unanswered regarding cell proliferation during appendage regeneration. How cell cycle dynamics change during regeneration compared with uninjured limbs, whether the cell cycle length is unique to individual regenerating organs, and whether the cell cycle is regulated differently during development versus regeneration are among some of the many unresolved issues. Previously, these questions were addressed using a combination of thymidine analogs such as EdU and BrdU, pHH3 antibody labeling, and mitotic figures. However, none of these methods provides information about multiple cell cycle stages and they are unable to be used for live imaging. FUCCI axolotls provide a powerful means with which to address these questions in detail while improving the existing toolbox for the study of cell cycle dynamics during tissue regeneration. Furthermore, a major advantage of using FUCCI sensors in the axolotl is the amenability of axolotl embryonic tissue to grafting, which can be used to label limb connective tissue, muscle cells, epithelium, Schwann cells, vasculature, neural stem cells, neural crest cells and teeth primordium (Epperlein et al., 2012; Kragl et al., 2009; Kragl and Tanaka, 2009; Nacu et al., 2013). Grafting FUCCI tissue onto white embryos will allow tissue-specific expression without the need for the generation of new transgenic animals with a tissue-specific promoter driving FUCCI expression.
Although FUCCI sensors in the axolotl are highly useful, some limitations exist. One issue of our transgenic line is the variable expression across animals and tissues. This is particularly obvious in adult animals, where some animals seem to strongly express only one fluorescent protein. Continuous breeding of the line to d/d mates should decrease variability across siblings. The use of EdU in FUCCI tissue also has small limitations. First, the number of mAG+/mCherry−/EdU− cells may be underrepresented and the number of mAG+/mCherry−/EdU+ may be overrepresented; after a 3 h pulse of EdU, cells labeled in late S phase will transition to G2 phase before collection. The severity of this issue can be reduced by collecting tissue sooner than 3 h after the EdU pulse, but this problem is theoretically always possible. Second, we observed the presence of mAG−/mCherry+/EdU+ and mAG−/mCherry−/EdU+ cell populations in our samples. Our quantification suggests that these populations are not highly abundant, and we speculate that they are detected as a result of the cells not expressing the FUCCI construct or DNA damage, as EdU is known to be incorporated into cells undergoing DNA repair (Verbruggen et al., 2014). An interesting finding in our study is the abundance of mAG−/EdU+ enterocytes in the axolotl gut (Fig. S1F-F″″). One explanation for this could be the rapid clonal expansion and migration of enterocytes from the intestinal crypt to the tip of the villi. This is supported by a decrease in EdU intensity at the tip of the villi, suggesting EdU dilution after division despite a relatively short 3 h pulse of EdU. More studies on the cell cycle length in the intestinal stem cell niche in the axolotl are required to elucidate this phenomenon.
In our study, we highlight the versatility of imaging FUCCI tissue with live imaging, whole-mount imaging and multimodal imaging. To our knowledge, we present the first real time movie of blastema formation in regenerating axolotl tail. The approaches used in this study will be helpful for other groups attempting to capture live images of appendage regeneration in real-time, but methods should be optimized. During blastema growth, we observed a large number of dying cells and cells sluffing from the blastema. After removing the larvae from the agarose, we also noticed that the blastema was misshapen. We suspect that this is not a true representation of tail blastema growth and is more likely a result of the blastema growing in the agarose. However, we cannot exclude the possibility of live cell extrusion (Chen et al., 2016; Eisenhoffer et al., 2012) or apoptosis. Using available programs for cell tracking, we also demonstrate the capability for tracking FUCCI cells during the regenerative process. Although images were taken every 30 min, we cannot be absolutely certain about the fate of some cells given this time interval without identification of specific cells. We also visualized FUCCI expression in whole-mount regenerating limbs with light-sheet fluorescence microscopy. We attempted to visualize these limbs in 3D after staining for EdU, but the EdU signal from the 647 channel bled into the mCherry signal. With better filtering, we expect to be able to perform 3D multiscale analysis of DNA synthesis in FUCCI tissue (Duerr et al., 2020).
Finally, we outline a method for cell characterization in transgenic tissue with multimodal imaging. These proof of concept experiments demonstrate that robust cell type information can be acquired from FUCCI tissue sections despite the endogenous mAG and mCherry signal. We foresee better cell type characterization via multimodal imaging with more rounds of FISH, multiple macromolecule analogs with unique click-it compatible functional groups (Duerr et al., 2020), and multiple primary antibodies raised in different species. One potential limitation of this method is the inefficiency of photobleaching FUCCI signal in large tissues. The spinal cord is an ideal organ for this analysis, as it can easily fit into a single 20X frame. Thus, photobleaching is contained to one single tile. If using larger tissue like a limb blastema, many more tiles may require photobleaching and at potentially a lower magnification, both of which will increase the time necessary to completely photobleach the FUCCI signal. Nevertheless, our method can be used for the robust characterization of cycling cells during tissue regeneration. We expect that these methods can also be applied in other transgenic reporter axolotl lines and other animal models.
MATERIALS AND METHODS
All transgenic animals were bred at Northeastern University, and all procedures and surgeries were approved by the Northeastern University Institutional Animal Care and Use Committee. Surgeries were performed while axolotls were anesthetized in 0.01% benzocaine. EdU was administered via intraperitoneal injection at 8.0 ng/g animal and samples were collected 3 h after injection.
Transgenesis was performed via I-SceI meganuclease digestion according to Khattak et al. (2009). Briefly, 1 µg of purified, CAG-FUCCI plasmid (Addgene #175266) was mixed in solution with 1 µl NEB Cutsmart buffer and 1 µl I-SceI enzyme, filled to a final volume of 10 µl with nuclease-free water to generate the FUCCI injection cocktail. Single-cell, d/d axolotl embryos were injected with 5 nl of FUCCI injection cocktail and grown to stage 45 for phenotype assessment. Owing to the β-actin promoter in CAG, the most intense FUCCI expression was observed in myomeric muscle of developing tails. Larvae with strong, ubiquitous expression were identified and grown to sexual maturity.
Histology and staining
Samples were fixed in 4% PFA overnight at 4°C, and after washing with 1×PBS three times for 5 min, samples were cryoprotected in 30% sucrose until equilibrated. Samples were then placed in OCT and frozen at −80°C. Frozen samples were sectioned with a cryostat to obtain 10 µm sections. Slides were then baked at 65°C for 15 min. Residual OCT was removed from slides by placing in water for 5 min at room temperature. From this step, the slides are ready for click-chemistry, IHC or FISH.
EdU labeling via click chemistry
For EdU detection, we used an Alexa-fluor 647 azide plus probe from clickchemistrytools.com (product number: 1482). The 1 ml click-it cocktail was made as follows: 888 µl 1×Tris, 10 µl 50 mM CuSO4 (0.5 mM final), 2 µl Alexa-fluor 647 azide plus (2 µM final) and 100 µl 100 mM sodium ascorbate (10 mM final). This cocktail was applied to slides for 30 min at room temperature.
Slides were incubated in blocking buffer (15 µl goat serum in 1 ml 1× PBS) for 30 min. Rabbit anti-pHH3 antibodies (Cell Signaling, 9701) were diluted in blocking buffer at 1:400 and applied to slides overnight at 4°C. Slides were washed three times for 5 min each with 1× PBS, and Alexa-fluor 647 anti-rabbit secondary antibodies (Invitrogen, A21245; 1:500 in 1× PBS) were applied to slides for 30 min at room temperature.
All of the following steps were conducted using RNase-free reagents. Slides were placed in 100% ethanol at room temperature for 1 h. Following three 5-min washes with 1× PBS, slides were prehybridized with hybridization buffer (Molecular Instruments) for 30 min at 37°C. Probe stocks for a particular transcript of interest were made to contain 1 µM of each oligo in 200 µl of RNase-free water. Probe sequences for Shh, B3Tub and Pax7 are provided in Table S3. Probe stocks were further diluted at 1:200 in hybridization buffer and applied to slides overnight at 37°C. Slides were washed with formamide wash buffer (Molecular Instruments) three times for 15 min at 37°C to remove unbound probe, then washed twice with 5× SSCT (20× saline sodium citrate with 0.1% Tween 20) for 15 min at room temperature. Amplification buffer (Molecular Instruments) was then applied to the slides for 30 min at room temperature. Fluorescent hairpins for each initiator (Molecular Instruments) were prepared by heating H1 and H2 hairpins to 95°C for 90 s. Hairpins were allowed to cool to room temperature in the dark, then diluted 1:50 in amplification buffer and applied to slides overnight at room temperature. Slides were then washed twice for 30 min with 5× SSCT at room temperature.
After these protocols, cell nuclei were stained with DAPI (2.86 µM) for 5 min at room temperature. Following a 5-min 1× PBS wash at room temperature, slides were mounted with SlowFade gold antifade mountant (Thermo S36936) and imaged using a Zeiss LSM 800 confocal microscope.
Live FUCCI imaging
Larvae used for live imaging were mounted in a 50×9 mm Petri dish in 0.3% low melt agarose diluted in 0.005% benzocaine. All live images were acquired using a Zeiss LSM 880 confocal microscope fitted with a humidification chamber to prevent sample desiccation. Larvae were imaged at 10× magnification. For live imaging of tail regeneration, we imaged two adjacent tiles to accommodate for growth during imaging at 10× magnification. Additionally, to accommodate cells moving in and out of the focal plane, we imaged four planes in the z-axis spanning 119.4 µm and merged these planes in a maximum intensity projection. To prevent photobleaching of the FUCCI probes, we used 1.0% laser power for each channel. Larvae were removed from agarose after imaging and placed in salamander housing water. Larvae were swimming and feeding 1 week after imaging with no visible signs of illness or distress.
EdU-pulsed FUCCI spinal cords were collected as outlined above. In the first round of multimodal imaging, we performed V3.HCR-FISH for Shh with 647 hairpins. The endogenous FUCCI signal and V3.HCR-FISH was then imaged. To photobleach the FUCCI signal, the 488 and 594 lasers were set to 100% laser power and were directed onto the spinal cord for 40 min. We found that the DAPI signal was weakened after this photobleaching, but was still present. The V3.HCR-FISH signal was sufficiently photobleached, but to ensure Shh probes were not amplified in the subsequent round of FISH, the slides were washed with 80% formamide four times for 15 min each at 37°C. The slides were washed in 5× SSCT twice for 15 min each at room temperature, prehybridized with hybridization buffer for 30 min at 37°C, and rehybridized with Pax7 and B3Tub probes for the second round of multimodal imaging. These probes were amplified with 647 and 488 hairpins, respectively. Slides were imaged, and probes were again removed with four 15 min washes of 80% formamide at 37°C. Slides were washed three times with 1× PBS, and the click-it cocktail outlined above was used with Alexa-fluor 647 azide plus probes for EdU labeling in the third round of multimodal imaging. Slides were imaged and treated with DNase I (NEB M0303) overnight at room temperature. DNase I was applied to slides without buffer, and enough was used to cover the entire section being imaged. The next day, we performed IHC with rabbit anti-B3TUB antibodies (Invitrogen, PA5-85639; 1:500) and applied anti-rabbit Alexa-fluor 647 antibodies the subsequent day. Slides were then imaged for the final round of multimodal imaging. Adobe Photoshop was used to align the images from each round onto the original DAPI image. All images were obtained with a Zeiss LSM 800 confocal microscope.
Whole-mount FUCCI imaging
To prepare whole mount-tissue, limbs were fixed in 4% PFA overnight at 4°C. Limbs were then washed with 1× PBS three times for 5 min, and dehydrated in an increasing methanol series (25% methanol/75% 1× PBS, 50% methanol/50% 1× PBS, 75% methanol/25% 1× PBS, each step for 5 min at room temperature), and stored in 100% methanol at −20°C for up to 6 months before imaging. Once ready to be imaged, the limbs were rehydrated in a decreasing methanol series (75% methanol/25% 1× PBS, 50% methanol/50% 1× PBS, 25% methanol/75% 1× PBS, each step for 5 min at room temperature) and washed once with 1× PBS for 5 min. The samples were washed three times with 1× PBST (PBS and 1% Triton X-100) for 5 min at room temperature. We found that a 90 min 0.5% trypsin treatment at room temperature with rocking improved light penetration of FUCCI samples without appreciable changes in intensity of mAG and mCherry. After trypsin treatment, samples were washed with deionized water three times for 5 min at room temperature. The limbs were then placed in 100% acetone for 20 min at −20°C. Afterwards, the samples were incubated in deionized water for 10 min at room temperature. Samples were again washed with 1× PBS three times for 5 min, mounted in 1.5% low melt agarose and refractive index matched with EasyIndex RI 1.465 (LifeCanvas Technologies) overnight at 4°C. Three-dimensional images were obtained using a Zeiss light-sheet Z1 microscope with Zen software. All post-processing for visualization was performed using Arivis software.
Cell dissociation, Hoechst staining and flow cytometry
Blastemas from the d/d strain or FUCCI animals were collected and pooled together in a six-well plate. The wound epithelium was not surgically removed. Blastemas were incubated on ice in 0.35 mg/ml Liberase for 20 min and were then incubated at room temperature with gentle agitation. Every 10 min, the tissue was manually dissociated with forceps by teasing the tissue apart. This was repeated until there was a sufficient single-cell suspension (checked under the microscope) while the wound epithelium remained intact. The wound epithelium was manually removed with forceps from the suspension. 1 ml of 80% PBS was added to the cell suspension prior to filtering in a 35 µm filter tube. The strainer was then washed with an additional 1 ml of 80% PBS.
The filtered single-cell suspensions were run on a BD FACSAria Fusion Cell Sorter (UMass Boston Flow Cytometry Core) using the 100 µm nozzle and the FSC 2.0 ND filter. The gates for the blastema cell population were set on the blastema cells from a d/d animal using forward and side scatter. Using the forward scatter and side scatter plot, the cell population was gated to separate it from debris and doublets. A sample of the gated population was sorted, and the presence of singlet cells was confirmed with fluorescence microscopy.
To identify the fluorescent negative blastema cells from the FUCCI animals, we first used blastema samples from d/d strain axolotls. To identify the red or green fluorescent populations, we first plotted each fluorescent channel versus forward scatter. Both plots yielded two discrete populations: a mAG−/mCherry−population, which overlapped with the previously identified fluorescent negative gate described above; and a fluorescent positive population. The mAG and mCherry fluorescent populations were each gated, and these gates were used to establish quadrant gates in the FITC (mAG) versus PE-Texas Red (mCherry) scatter plot. The quadrant gates were added to quantify the mAG−/mCherry+ (G1 phase), mAG+/mCherry− (S, G2 and M phases) and mAG+/mCherry+ (G1 to S transition) populations. Gates for fluorescent populations were also confirmed by sorting and validating cell populations with fluorescence microscopy.
To analyze the DNA content in each of the FUCCI populations, samples were stained with Hoechst for DNA content analysis. Hoechst was added to the single-cell suspension at a final concentration of 1.62 µM for 10 min at room temperature before filtering through a cell strainer. The lack of clear separation between the 2N and 4N peaks in the Hoechst channel required the use of the red and green fluorescent-negative population to gate the 2N population in the Hoechst channel. The 4N gate was set around the cell populations with stronger Hoechst staining than the 2N gate. Using these parameters, 84% of the mAG−/mCherry− fluorescent population was in the 2N gate, and 65% of the mAG+/mCherry+ cells were in the 4N gate.
FUCCI+ cells were quantified either manually in Adobe Photoshop or with Cellpose (Stringer et al., 2021) combined with custom Fiji scripts (Schindelin et al., 2012), which have been deposited in GitHub (https://github.com/Monaghan-Lab/FUCCI-macros). Fiji scripts for quantification of mAG/mCherry intensity changes during tail regeneration have also been deposited in GitHub (https://github.com/Monaghan-Lab/FUCCI-macros). Limb blastema amputation plane to mCherry+ muscle line measurements were made using the InteredgeDistance macro on Fiji. Data processing and statistical analysis were conducted in Microsoft Excel and Matlab.
The authors thank Jackson Griffiths for assisting with transgenic animal care during the COVID-19 pandemic lockdown, Ester Comellas for assistance with data analysis, both Guoxin Rong and Alex Lovely for imaging assistance, and Malcolm Maden for early conceptual discussions on the project. Images were obtained from the Northeastern University Chemical Imaging of Living Systems core. We thank the Institute for Chemical Imaging of Living Systems at Northeastern University for consultation and imaging support. Non-transgenic animals were obtained from the Ambystoma Genetic Stock Center, which is funded by the National Institutes of Health (P40OD019794).
Conceptualization: T.J.D., A.W.S., J.R.M.; Methodology: T.J.D., A.V., J.R.M.; Software: A.V.; Validation: T.J.D.; Formal analysis: T.J.D., A.V.; Investigation: T.J.D., E.K.J., K.M.W., J.R.M.; Resources: C.D.M., J.R.M.; Data curation: T.J.D.; Writing - original draft: T.J.D., K.M.W., C.D.M., J.R.M.; Writing - review & editing: T.J.D., E.K.J., A.V., A.W.S., C.D.M., J.R.M.; Visualization: T.J.D., E.K.J., A.V.; Supervision: J.R.M.; Project administration: J.R.M.; Funding acquisition: J.R.M.
The work from this paper was funded by the National Institutes of Health (R01HD099174) and by the National Science Foundation (1558017 and 1656429). Funding for A.V. and E.K.J. was partially provided by the National Science Foundation (REU grant 1757443). Deposited in PMC for release after 12 months.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199637.
The authors declare no competing or financial interests.