Fgfs control homeostatic regeneration in adult zebrafish fins

The capacity for organ regeneration is remarkably elevated in some non-mammalian vertebrates such as urodele amphibians and teleost fish. Together, these species regenerate amputated appendages and jaws, resected heart muscle, depleted sensory hair cells, damaged retinae and brain,dissected lenses, transected spinal cord and regions of intestine. The teleost zebrafish represents a unique example of a highly regenerative vertebrate model system that is amenable to genetic approaches, making it well suited to illuminate how complex tissue regeneration occurs at the molecular level. One of the most spectacular examples of regeneration in adult zebrafish is their rapid and virtually indelible renewal of amputated fins, complex tissues comprised of segmented bone, connective tissue, blood vessels, nerves and epidermis. Fin regeneration is divided into three stages - wound healing,blastema formation, and regenerative outgrowth(Akimenko et al., 2003; Poss et al., 2003; Stoick-Cooper et al., 2007a). First, the amputation injury is healed through migration of epidermal cells,covering the wound within an hour of trauma. Formation of the regeneration blastema, a mass of undifferentiated, proliferative mesenchymal cells, is the second and defining stage of regeneration. Here, presumptive blastemal cells are stimulated to disorganize, migrate distally and accumulate within 1-2 days of injury. It is currently unclear whether the blastema consists of a homogeneous population of multipotent progenitor cells, or heterogeneous subpopulations. During the final stage called regenerative outgrowth, a repeated series of proliferation, patterning and differentiation events in blastemal tissue leads to deposition of new bone-secreting scleroblasts and distal addition of all lost bone segments within ∼2 weeks.

Multiple studies in recent years have identified key molecular regulators of blastemal formation and function. Several different approaches revealed that signaling by fibroblast growth factors (Fgfs) is crucial for regeneration(Lee et al., 2005; Poss et al., 2000; Tawk et al., 2002; Thummel et al., 2006; Whitehead et al., 2005). The ligand fgf20a, which was found in a genetic screen for temperature-sensitive regeneration mutants, is required for normal morphogenesis of the regeneration epidermis and for mesenchymal proliferation during blastema formation (Whitehead et al., 2005). During regenerative outgrowth, the Fgf receptor (Fgfr) fgfr1, as well as Fgf target genes mkp3 (dusp6 -Zebrafish Information Network), sef (il17rd - Zebrafish Information Network) and spry4, are expressed in blastemal mesenchyme and in the surrounding basal epidermal layer. As regeneration proceeds, Fgf signaling tightly controls the amount of blastemal proliferation and the rate of growth, resulting in different regenerative rates dependent on the proximodistal level of amputation (Lee et al., 2005; Poss et al.,2000). Other studies have shown that a suite of signaling molecules, such as sonic hedgehog (Shh), bone morphogenetic proteins,activin-βA, and canonical and noncanonical Wnts, influences blastemal proliferation and patterning during regenerative outgrowth(Jazwinska et al., 2007; Laforest et al., 1998; Quint et al., 2002; Smith et al., 2006; Stoick-Cooper et al.,2007b).

Vertebrate organs generally exhibit two forms of regeneration: facultative and homeostatic. Facultative regeneration describes mechanisms that are activated by stimuli like amputation or chemical injury: following the initial trauma, progenitor and/or structural cells near the injury site proliferate to replace dead or lost tissue. By contrast, homeostatic regeneration refers to regular replacement of cells lost through apoptosis, daily wear and aging(Jones and Wagers, 2008). Interestingly, surveys of regenerative capacity among mammalian organs have found that the capacity of an organ for facultative regeneration often correlates positively with its baseline level of cell turnover(Rando, 2006). For instance,blood and skin undergo frequent cell loss and replacement through the activity of self-renewing stem cells, and use similar processes to quickly regenerate after injury (Blanpain and Fuchs,2006; Scadden,2006). Conversely, the mammalian brain and heart possess low levels of cellular turnover, and, despite evidence for resident stem cells(Alvarez-Buylla and Lim, 2004; Beltrami et al., 2003; Laugwitz et al., 2005), there is little or no regeneration after major injury. Furthermore, facultative regenerative capacity in mammalian organs tends to decrease with age, a phenomenon observed in concert with age-dependent reductions in the frequency of homeostatic structural cell or progenitor cell proliferation(Janzen et al., 2006; Krishnamurthy et al., 2006; Molofsky et al., 2006). These correlations indicate that common cellular and molecular mechanisms are responsible for recurrent cell replacement and injury-induced regeneration in many tissues. However, neither the vigor nor the mechanisms of homeostatic regeneration have been examined in complex tissues that regenerate through a blastema-based mechanism, such as urodele or teleost appendages.

In this study, we tested the idea that the molecular pathways that control blastema formation and function during regeneration in amputated zebrafish fins have additional homeostatic functions in the absence of injury. We found that long-term inhibition of Fgfrs in uninjured zebrafish led to the progressive loss of distal fin structures, revealing homeostatic maintenance of fins by this pathway. Homeostatic regeneration was characterized by low-level expression of several mediators of facultative regeneration,including shh, msxb and mkp3, in areas of cell proliferation and apoptosis. Using a conditional mutant strain, we found that Fgf-dependent homeostatic regeneration is mediated at least in part by the specific ligand Fgf20a. Taken together, our findings reveal robust new requirements for Fgfs in the day-to-day homeostatic preservation of zebrafish appendages, and have implications for why elevated regenerative capacity has been selectively preserved in some vertebrate species.

Wild-type, transgenic and mutant zebrafish of the Ekkwill strain, or hybrids between Ekkwill and the related AB strain, were used in all experiments. Animals were between 4 and 12 months of age. mps1mutants, fgf20a mutants, and transgenic hsp70:dn-fgfr1 and shh:EGFP fish have been described previously(Lee et al., 2005; Poss et al., 2002a; Shkumatava et al., 2004; Whitehead et al., 2005). Heterozygous hsp70:dn-fgfr1 fish and wild-type clutchmates received daily heat shocks using an electric heater to raise the water temperature from 26°C to 38°C, as described previously(Lee et al., 2005). Homozygous mps1 and fgf20a mutants, and wild-type controls, were raised to adulthood at 25°C before transfer to aquaria with recirculating water heated to 33°C. Homozygous mps1 mutants were identified by phenotyping for regenerative defects at 33°C, and then allowed at least 1 month to regenerate at 25°C before use in homeostasis experiments. Some zebrafish heterozygous for mps1 mutations also showed loss of distal fin structures in our experiments (data not shown). For fin amputations,∼50% of the caudal fin was removed using a razor blade and animals were allowed to regenerate for 3 days at 33°C or 3-4 days at 25°C.

Fish were anesthetized in 0.1% tricaine and the fins were imaged at various timepoints. Imaging software (Openlab) was used to measure the length of the two rays flanking the central-most ray, from the end of the most proximal visible ray segment to the distal tip of the ray. The two values were then averaged to obtain a value for each fin. Each value was divided by the average fin length at day 0 (for the same group) to give a normalized (percentage of starting length) value of fin length. Central rays were chosen because they are the most protected from any potential spontaneous injury, although such injuries rarely occurred in our studies (see Fig. S1 in the supplementary material), and because the degree of tissue loss often varied between the two lobes of the fin, confounding statistical analysis. Tests of statistical significance were performed using Student's t-test, with two-tailed distribution assuming unequal variance. At least eight fins were assessed at each timepoint for each group.

Fish were separated into groups of four or five fish and placed in a 1 l tank so that individual fish were recognizable over the course of the experiment. Every 2-3 days, fins were imaged on a dissecting microscope, over a total of 14-24 days. A low-magnification image of the whole fin and a series of high-magnification images of groups of fin rays were acquired for each fin.

To visualize scleroblasts in tissue sections, fins were fixed in paraformaldehyde (PFA) overnight at 4°C and cryosectioned. Fin sections were then stained with the monoclonal zns-5 antibody as described(Johnson and Weston, 1995; Poss et al., 2002b). For whole-mount visualization of scleroblasts, the monoclonal zn-3 antibody was used, which marks the scleroblast cell membrane (A.A.W. and K.D.P.,unpublished). Fins were fixed in Carnoy's solution overnight at 4°C, and stained as described (Newmark and Sanchez Alvarado, 2000; Poss et al.,2002a).

For comparison of cell death in the distal and proximal regions of the fin,wild-type fins were fixed and cryosectioned, and slides were dried overnight at room temperature. Slides were then incubated in PBS for 30 minutes at 37°C. DNaseI-treated slides were used as a positive control. Slides were transferred to 0.3% Triton X-100 (Sigma) in PBS for 10 minutes, and then covered with 150 μl of 1× TdT buffer (Invitrogen) for 5 minutes. The buffer was then removed and replaced with 150 μl of 1× TdT buffer containing 0.3 U/μl of TdT enzyme (Invitrogen) and 8 μM Biotin-14-dCTP(Invitrogen). Slides were incubated at 37°C for 1 hour, followed by termination in stop buffer (300 mM NaCl, 30 mM sodium citrate) for 15 minutes at room temperature. Slides then rinsed three times in PBS and covered with 20μg/ml Texas Red Streptavidin (Vector Laboratories) in PBS (pH 8.2), and incubated in the dark at room temperature for 30 minutes. Slides were then washed four times in PBS and coverslipped using Vectashield with DAPI to stain nuclei.

Five different sections were imaged at distal and proximal regions of each fin (n=10 fins). Distal regions represent approximately 350 μm at the distal end of the fin, and proximal samples represent tissue ∼700μm to 1050 μm from the distal end of the fin. This length was chosen because it is the length of a frame at 20× magnification using our imaging equipment. Epidermal and mesenchymal regions were carefully outlined in each image, and the area quantified using Openlab software. Then,TUNEL-positive nuclei, co-stained with DAPI, were counted by hand within these regions. The areas and TUNEL-positive cell counts from these 5 sections were summed, giving each animal four indices: distal epidermal, distal mesenchymal,proximal epidermal and proximal mesenchymal cell death.

For comparison of proliferation in distal and proximal regions of the fin,fish were allowed to swim for 24 hours in a 50 μg/ml solution of bromodeoxyuridine (BrdU) in fish water. After collection, fins were fixed,cryosectioned and stained as described(Poss et al., 2002b). For each fin, five different sections were analyzed for both distal and proximal samples as described above. A total of eight fins were analyzed for the number of BrdU-positive cells per area of epidermal or mesenchymal tissue.

BrdU incorporation was also analyzed in wild-type or hsp70:dn-fgfr1 animals given 14 days heat shock followed by 5 days recovery at room temperature. For these experiments, a 2.5 mg/ml solution of BrdU was injected intraperitoneally 2 hours before fin collection. Fins were fixed in Carnoy's solution overnight at 4°C, and stained as described(Newmark and Sanchez Alvarado,2000; Poss et al.,2002a). A total of 10 wild-type and eight hsp70:dn-fgfr1fins were analyzed.

Whole-mount in situ hybridization was performed as described previously(Poss et al., 2000), using digoxigenin-labeled probes for mkp3, msxb, fgf20a and mps1(ttk - Zebrafish Information Network)(Akimenko et al., 1995; Lee et al., 2005; Poss et al., 2002a; Whitehead et al., 2005). Wild-type and transgenic fins were hybridized and developed simultaneously. Section in situ experiments were performed as described, using mkp3and msxb probes (Poss et al.,2002b). The msxb probe was developed overnight, as is standard in our laboratory (∼16 hours); for mkp3, the exposure time was shortened to ∼4 hours to limit nonspecific staining.

RNA samples were prepared from uninjured fins and regenerating fins using TRI Reagent (Sigma) according to the manufacturer's protocol. Total RNA (5μg) was used for reverse transcription reactions using Superscript III and Oligo-dT₂₀ (Invitrogen) according to the manufacturer's protocol. PCR reactions, consisting of 2 minutes at 94°C, 26-30 cycles of 94°C for 20 seconds, 52°C for 20 seconds, 72°C for 45 seconds and a final 72°C extension for 5 minutes were performed in an Eppendorf Mastercycler,and samples were run on 2% agarose gels.

To test the extent to which facultative mechanisms also support homeostatic regeneration of zebrafish appendages, we examined requirements for Fgfs,robust regulators of blastema formation and function in a variety of organisms and systems. We used a transgenic strain in which blastemal proliferation can be inducibly blocked by heat-induced expression of a transgenic dominant-negative Fgfr construct (hsp70:dn-fgfr1), slowing or blocking regeneration of amputated fins(Lee et al., 2005; Yin et al., 2008). Adult transgenic hsp70:dn-fgfr1 fish and wild-type clutchmates were given a strong, daily heatshock regimen for 30 or 60 days, after which animals were imaged to assess caudal fin length and morphology. Surprisingly, these conditions caused rapid, easily visible loss of distal fin tissue in transgenic animals (Fig. 1A). Transgenic fins often terminated with damaged bone and an excess of epidermal tissue, suggesting recent injury and epidermal repair(Fig. 1B). Wild-type clutchmates occasionally showed similar morphology (without tissue loss), but only in isolated rays rather than the entirety of the fin (data not shown). The degree of damage was variable in hsp70:dn-fgfr1 outer rays but generally consistent within the inner rays; therefore, we quantified the lengths of centrally located rays and found reductions of fin length of∼10% after 30 days, and ∼17% after 60 days. Under identical conditions, wild-type clutchmates maintained their fin length(Fig. 1C). Isolated hsp70:dn-fgfr1 animals, removed from possible aggressive interactions with other fish, also displayed progressive loss of fin structures (data not shown). Furthermore, similar tissue loss also occurred in dorsal, anal, pelvic and pectoral fins, indicating that Fgf signaling is required for homeostasis in all fin types (Fig. 1D; data not shown). If Fgf signaling was restored to these animals for 30 days following long-term Fgfr inhibition, the majority of fin rays were able to recover lost structures (see Fig. S1 in the supplementary material).

We examined transgenic fin tissues closely to identify the underlying pathology responsible for atrophy. One explanation was the inability during extended Fgfr inhibition to regenerate tissues lost through minor injuries that occur normally in zebrafish fins. In this model, zebrafish experience regular bone fragmentation and loss, recurrently countered by activation of regenerative mechanisms. To examine this possibility, we observed and imaged wild-type fin rays at high magnification every 2-3 days over a period of 2-3 weeks. We found that events of bone loss and regeneration occur very rarely in our animal facility (two events in 13 fish over 14-24 days; see Fig. S2 in the supplementary material). Although not fully excluding a model of recurring injury, these observations indicate that visible injuries to the fin are normally too rare or mild to account for the consistent degeneration we observed during Fgfr inhibition.

Interestingly, upon examining fins at high magnification after 30 days of Fgf inhibition, we found that over half of the transgenic fins exhibited swelling, discoloration and/or separation or slipping of rays at the intersegmental joints (16/30 fins, Fig. 2A). Joint swelling or segment separation was only occasionally observed in wild-type clutchmates (2/30 fins), and never with the degree of severity exhibited by transgenics. This suggested that pathology or weakness at segment joints contributes to major tissue loss in hsp70:dn-fgfr1fins. We examined the joints more closely by confocal microscopy and by histology to identify disturbances in cellular organization. The zebrafish fin lepidotrichia consist of two bony, facing hemirays, which are initially formed by mineralization of bone matrix secreted by the flattened scleroblasts that encase them. At the intersegmental joints between lepidotrichial segments,there exist clusters of scleroblasts with a different rounded morphology. In projections of confocal slices and in tissue sections, these rounded scleroblasts appear as a small bulge in the intraray mesenchyme around the joints (Fig. 2B,C). Scleroblasts in regenerating fins also have a similar rounded morphology(Fig. 2C). In confocal projections and sections through hsp70:dn-fgfr1 fin rays, the bulge at the intersegmental joints was often enlarged, with large numbers of rounded scleroblasts protruding deep into the surrounding mesenchyme(Fig. 2B,C). We observed this hypertrophy in areas of severe joint dysmorphology(Fig. 2C, bottom left), as well as in transgenic rays that lacked major anatomical changes(Fig. 2B, middle and bottom right). Our observations indicate that one way in which Fgf signaling maintains fin integrity is through control of scleroblast activity and/or patterning, particularly at intersegmental joints.

The fin atrophy exposed by Fgfr blockade indicated vigorous homeostatic regeneration; thus, we searched for regions of cell turnover and activated developmental programs in uninjured fins. Previous studies have described cell proliferation in the uninjured zebrafish caudal fin, reporting qualitatively higher levels of proliferation at the distal fin structures and lower levels in more proximal regions. This graded proliferation is purported to reflect growth; specifically, coordinated bursts or saltations of proliferation that periodically add new ray segments as part of adult indeterminate growth(Goldsmith et al., 2003; Iovine and Johnson, 2000; Nechiporuk and Keating, 2002). We quantified cell proliferation in distal and medial regions of uninjured fins using assays for BrdU incorporation(Fig. 3A). Proliferation was low compared with that seen during regeneration of amputated structures. Nevertheless, our data confirmed significantly higher BrdU incorporation indices in distal structures. Epidermal BrdU incorporation was ∼43% higher in the most distal 350 μm of the fin versus an identical length of medial structures, whereas BrdU incorporation in mesenchymal tissue, which includes connective tissue cells and scleroblasts, was ∼197% higher in distal structures (Fig. 3B,D). Although proliferation was highest at the distal ray tip, proliferating cells within both the epidermal and mesenchymal compartments were not restricted to a specific cell type or a concentrated area as prominent as the regeneration blastema.

We postulated that these graded differences in proliferation indices in uninjured fins might reflect graded cell turnover, in addition to contributions to distal organ growth. To test this idea, we assessed apoptosis by TUNEL staining. Both mesenchymal and epidermal compartments all along the proximodistal axis of the fin contained low numbers of apoptotic cells. However, each of the mesenchymal and epidermal TUNEL indices was significantly higher in distal fin tissue than in more medial tissue(Fig. 3C,D). That rates of TUNEL-detectable cell apoptosis were higher in more proliferative areas of the fin suggests that the ongoing cell proliferation in distal fin structures represents, at least in part, a homeostatic response to counter higher levels of cell death.

Next, we searched for evidence of developmental programs associated with these turnover events, paying particular attention to programs known to regulate regeneration of amputated fins. We first took advantage of an transgenic EGFP reporter strain that marks expression of shh, a reported blastemal mitogen and patterning factor synthesized in regeneration epidermis adjacent to the blastema(Laforest et al., 1998; Quint et al., 2002; Shkumatava et al., 2004). We found that this strain visualizes shh expression during regeneration in a manner consistent with published in situ hybridization results. Moreover, shh was consistently detectable in similar epidermal expression domains in the distal-most ∼100 μm of uninjured fins(Fig. 3E). We similarly examined two other markers: (1) msxb marks cells throughout the regeneration blastema as well as regenerating scleroblasts, and was recently shown using electroporation of antisense morpholinos to be essential for regenerative growth (Akimenko et al.,1995; Thummel et al.,2006); and (2) mkp3 is an Fgf target gene that is induced in the blastema as well as cells of the basal epidermal layer encasing the blastema, with expression levels that correlate with the rate of regenerative outgrowth (Lee et al., 2005). Semi-quantitative RT-PCR demonstrated that both of these markers are expressed in the uninjured fin, albeit at much lower levels than that seen after amputation (see Fig. S3 in the supplementary material). By in situ hybridization of tissue sections, we detected faint expression of both of these molecules at the distal fin tip. Interestingly, detectable expression was restricted to domains similar to those occupied by these molecules during regenerative outgrowth: mkp3 was expressed in cells of the distal epidermis and mesenchyme, while msxb expression was restricted to mesenchymal cells, suggesting that these molecules have similar functions in amputated and uninjured fins (Fig. 3F). Specific localization of these regeneration mediators suggested homeostatic involvement in the cell turnover and maintenance functions we had identified.

Do cell proliferation and low-level shh, msxb and Fgf signaling/mkp3 expression in uninjured fins reflect ongoing homeostatic events? To test whether these events rise in the face of a greater homeostatic obligation, we gave transient periods of Fgfr inhibition with the intention that it might `prime' fins for a burst of proliferation and marker expression after Fgf signaling is restored. In these experiments, a 7- or 14-day protocol of daily heatshocks was given to hsp70:dn-fgfr1 and wild-type clutchmates, followed by a 5-day recovery period at room temperature(Fig. 4A). We saw no significant difference in TUNEL-positive cells in transgenic fins versus wild-type fins during the heatshock period, suggesting little effects of Fgfr inhibition on apoptosis, as well as no significant differences in mesenchymal BrdU incorporation (data not shown). The latter observation might reflect a role for Fgfs in proliferation of just a subset of cell types within the uninjured fin. Alternatively, our assay may be limited by variability caused by very low baseline numbers of proliferating cells or the possibility of missing saltatory bursts of proliferation.

By contrast, following the recovery period from heatshock, proliferation was markedly increased in distal fin regions of transgenic, but not wild-type animals, indicative of an enhanced homeostatic response(Fig. 4B). Furthermore,expression of mkp3 and msxb could be easily visualized by whole-mount in situ hybridization in transgenic fins, but not wild-type fins,during the recovery period of restored Fgf signaling(Fig. 4C). These changes would not be expected if marker expression and cell proliferation were a purely ontogenetic or saltatory growth signature of adult fins. Together, these data reveal molecular and cellular indicators of a responsive program of homeostatic regeneration in uninjured zebrafish fins.

Although the above experiments demonstrate that a long-term blockade of Fgfrs causes fin regression, it remained possible that such a strong,animal-wide manipulation disrupts fins indirectly through effects on other organs. For example, extrinsic blood-borne factors that change with age are known to regulate the regenerative capacity of skeletal muscle satellite cells(Conboy et al., 2005). To address this possibility and also refine our analysis of Fgf signaling during homeostasis, we focused on fgf20a, an Fgf ligand with functions believed to be specific for regeneration of amputated adult fins. Incubation of developing fgf20a embryos at the restrictive temperature of 33°C has little effect on ontogenetic development, whereas incubation at this temperature arrests adult fin regeneration during blastema formation(Whitehead et al., 2005). Semi-quantitative RT-PCR revealed that fgf20a is expressed at low levels in the uninjured fin (see Fig. S2 in the supplementary material), and in situ hybridization of fins from primed hsp70:dn-fgfr1 animals indicated that fgf20a is a member of a homeostatic response program that includes msxb and mkp3(Fig. 5A). These expression data suggested that fgf20a might mediate Fgf-dependent homeostatic regeneration.

To test whether fgf20a is essential for appendage homeostasis, we placed adult fgf20a mutants at 33°C for 30-60 days and assessed fin size and integrity. Whereas wild-type controls maintained fin length at 33°C in these experiments, fgf20a zebrafish displayed a progressive loss of distal fin structures that was quantitatively very similar to results observed with hsp70:dn-fgfr1 animals(Fig. 5C). Quantification of central ray length revealed that fgf20a mutants lost ∼19% of their fin tissue after 30 days at the restrictive temperature, and roughly one-third of their fin length (∼35%) after 60 days(Fig. 5D). Unlike hsp70:dn-fgfr1 zebrafish, joint morphology appeared normal by gross visual inspection, and tissue loss at outer fin rays was less severe. This may be related to different strengths of genetic interventions, different conditions for removing Fgf functions, genetic redundancy and/or functions specific to Fgf20a apart from other Fgf ligands. In any case, our data indicate that signaling by Fgf20a is essential to maintain zebrafish appendages, accounting for at least part of the Fgf dependency of homeostatic regeneration.

To test whether other known regulators of the blastema show similar expression characteristics and requirements during homeostasis, we examined mps1, a dual-specificity kinase essential for the spindle checkpoint(Fisk and Winey, 2004). mps1 expression is induced in the newly formed blastema and colocalizes with highly proliferative blastemal cells during regenerative outgrowth. Like fgf20a, mps1 was shown to be essential for regeneration by positional cloning of a temperature-sensitive mutation(Poss et al., 2002a). RT-PCR assays indicated low-level mps1 expression in the uninjured fin (see Fig. S2 in the supplementary material), and, like fgf20a, mps1expression was induced by a 14-day block of Fgf signaling and 5 days of recovery in hsp70:dn-fgfr1 animals(Fig. 5B). When mps1mutant zebrafish were placed at the restrictive temperature of 33°C for 30 or 60 days, the animals showed severe distal tissue loss. Measurement of the central rays revealed a ∼17% loss of tissue by 30 days at 33°C, and∼36% after 60 days (Fig. 5C,D). Thus, Mps1, a kinase essential for proliferation of blastemal cells after fin amputation, is also necessary to maintain tissue in intact fins. Taken together, our data suggest that factors that enable facultative regeneration are generally required for homeostatic regeneration within uninjured fins.

We conclude that the Fgf-dependent developmental machinery responsible for blastema-based facultative regeneration of zebrafish fins also mediates homeostatic events that maintain existing fin ray structures. Most notably,rapid loss of distal appendage tissue occurred after multiple genetic disruptions of programs that form and maintain the regeneration blastema. Two independent methods to block Fgf signaling had striking effects on tissue maintenance, pinpointing a requirement for Fgf20a in preserving complex appendage tissue. The kinase Mps1 is also crucial for fin homeostasis. Our phenotypic data suggest that Fgf20a functions in maintenance of distal fin regions where cell turnover and expression of msxb, shh and mkp3 are most visible. We also observed effects of Fgfr blockade on joint anatomy that indicate additional functions of this signaling pathway.

Although facultative and homeostatic regeneration appear to have many similarities, there are also interesting differences. For example, fgf20a mutants have a mostly penetrant defect in injury-induced fin regeneration at 25°C (Whitehead et al., 2005), despite developing normally to adulthood and maintaining fins grossly normally at this temperature. This suggests that there is a requirement for some amount or function of Fgf20a that is met during homeostatic regeneration but not facultative regeneration. Similarly,redundancies may be present in one type of regeneration and absent in another. Other differences may relate to restrictive programs with greater presence in intact fins. For example, recent work has found that signaling by epidermally synthesized Wnt5b has an inhibitory role on blastemal proliferation(Stoick-Cooper et al., 2007b). Candidates for analogous factors in intact fins include several miRNAs that were recently shown to have higher levels in uninjured fins than in fins regenerating after amputation. At least one of these miRNAs, miR-133, also displayed functional properties of a regenerative brake(Yin et al., 2008). Understanding the balance between permissive and restrictive factors in injured and uninjured appendages is likely to be important in unraveling seminal issues in regeneration.

One of the most fascinating aspects of appendage regeneration is positional memory, the ability of the limb or fin stump to recognize and restore only those structures lost by injury. Positional memory is thought to be based on a gradient of some determinant(s) existing in uninjured tissue or quickly established after amputation. Recently, we found that the amount of Fgf signaling established after amputation is graded along the proximodistal axis,with higher amounts in more proximal tissue and lower amounts distally(Lee et al., 2005). Greater Fgf signaling positively impacts blastemal proliferation and regenerative rate, leading to more rapid outgrowth in proximal regenerates. Similarly, as regeneration proceeds gradually to completion, amounts of Fgf signaling gradually wane.

Our findings here indicate that developmental signaling, including Fgf signaling, is not inactivated after restoration of lost structures; rather, a modicum is maintained at low levels in the distal tips of intact fins. When the capacity to maintain this Fgf signaling is experimentally blocked, tissue loss occurs, revealing an essential role in homeostatic equilibrium. Based on these earlier and current observations, we postulate that weak regenerative presence at the distal tips of intact fins may in fact be a marker or determinant of far-distal positional identity. That is, uninjured fins maintain size in part due to a level of Fgf signaling that precisely opposes ongoing cell death. By contrast, an amputated fin will initiate a position-dependent boost of Fgf signaling for structure-restoring growth. Regeneration then culminates when Fgf signaling decreases to an amount that no longer procures a net gain in growth vis à vis ongoing cell death. Although this is an attractive model, what requires characterization is the signal(s) that determines position and sets the appropriate amount of Fgf signaling.

Why is the capacity for appendage regeneration, or organ regeneration in general, distributed unequally among vertebrate species? The selective advantages for non-mammalian regenerative events are not uniformly obvious. For example, the capacity for tail regeneration in lizards facilitates repeated use of the anti-predatory tactic of autotomy, and also is thought in some species to reduce potential reproductive, social and locomotor costs(Clause and Capaldi, 2006). By contrast, there is no immediate explanation for the capacity of adult non-mammalian vertebrates to regenerate resected cardiac muscle(Poss et al., 2002b).

Our current study, supported by previous experiments(Nechiporuk and Keating,2002), suggests that one contributing factor behind the high regenerative capacity of adult zebrafish fins is the particularly dynamic nature by which they are actively maintained. That is, the preservation of capacity to regenerate patterned fin structures after major injury might be the evolutionary consequence of a more crucial role for regenerative mechanisms to balance regularly day-to-day cell loss and maintain existing tissue. Accordingly, it is interesting to speculate that teleosts and urodeles possess a wider array of regenerative tissues than mammals, in part, because of greater cell turnover among those organs. This might also be related to the capacity for indeterminate growth in many of these species, although the newt does not grow throughout its adult life. Supporting this idea, we have found that the adult zebrafish heart, unlike the more static mammalian heart,actively adds cardiac cells during adult animal growth and size maintenance(Wills et al., 2008). Similarly, adult teleost CNS structures like the highly regenerative retina and the brain show unusually high basal rates of neurogenesis(Grandel et al., 2006; Otteson and Hitchcock,2003).

Notably, the severe fin regression phenotypes we observed after genetic manipulations in zebrafish are highly reminiscent of those seen recently after various gene knockdown experiments in the classic invertebrate model system for blastema-based regeneration: freshwater planarians. Rapid facultative and homeostatic regeneration in planarians are based on stem cell-like neoblasts,the sole proliferative cell type responsible for renewal of all structural cells (Birnbaum and Sanchez Alvarado,2008). In those studies, long-term RNA interference of individual genes frequently had similar effects on facultative and homeostatic regeneration; that is, regeneration could be blocked after animal bisection,while fatal `curling' or progressive changes in adult pattern occurred within weeks of gene perturbation in intact animals(Cebria et al., 2007; Cebria et al., 2002; Cebria and Newmark, 2005; Reddien et al., 2007; Reddien et al., 2005). It will be interesting to compare among different vertebrate species and organs the extent to which the capacity for injury-induced regeneration correlates with the activity of ongoing homeostatic regeneration maintained by Fgf ligands and other signaling pathways.

We thank G. Whitehead and E. Lien for sharing fgf20a mutants; J. Moss for shh:EGFP fish; J. Bumgardner, O. Ighile and C. Wheeler for excellent zebrafish care; J. Holdway for help with imaging; the Zebrafish International Research Center for antibodies; and K. Iovine, Y. Lee, P. Newmark, E. Tanaka, V. Yin and Poss laboratory members for helpful comments on the manuscript. This work was supported by grants to K.D.P. from NIGMS, NHLBI,Whitehead Foundation, March of Dimes and Pew Charitable Trusts.