Conditional targeted cell ablation in zebrafish: A new tool for regeneration studies


  • Silvia Curado,

    1. Department of Biochemistry and Biophysics, Programs in Developmental Biology, Genetics and Human Genetics, Liver Center, Diabetes Center and the Cardiovascular Research Institute, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Ryan M. Anderson,

    1. Department of Biochemistry and Biophysics, Programs in Developmental Biology, Genetics and Human Genetics, Liver Center, Diabetes Center and the Cardiovascular Research Institute, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Benno Jungblut,

    1. Department of Biochemistry and Biophysics, Programs in Developmental Biology, Genetics and Human Genetics, Liver Center, Diabetes Center and the Cardiovascular Research Institute, University of California at San Francisco, San Francisco, California
    Search for more papers by this author
  • Jeff Mumm,

    1. Department of Anatomy and Neurobiology, Washington University, St. Louis, Missouri
    Current affiliation:
    1. Luminomics, Inc. 1508 South Grand Blvd., St. Louis, MO 63104
    Search for more papers by this author
  • Eric Schroeter,

    1. Department of Anatomy and Neurobiology, Washington University, St. Louis, Missouri
    Current affiliation:
    1. Department of Biology, Loyola University Chicago, Quinlan Life Sciences Center 317, 6525 N. Sheridan Road, Chicago, IL 60626
    Search for more papers by this author
  • Didier Y.R. Stainier

    Corresponding author
    1. Department of Biochemistry and Biophysics, Programs in Developmental Biology, Genetics and Human Genetics, Liver Center, Diabetes Center and the Cardiovascular Research Institute, University of California at San Francisco, San Francisco, California
    • Department of Biochemistry and Biophysics, Programs in Developmental Biology, Genetics and Human Genetics, Liver Center, Diabetes Center and the Cardiovascular Research Institute, University of California, San Francisco, 1550 Fourth Street, San Francisco, CA 94158-2324
    Search for more papers by this author


Conditional targeted cell ablation in zebrafish would greatly expand the utility of this genetic model system in developmental and regeneration studies, given its extensive regenerative capabilities. Here, we show that, by combining chemical and genetic tools, one can ablate cells in a temporal- and spatial-specific manner in zebrafish larvae. For this purpose, we used the bacterial Nitroreductase (NTR) enzyme to convert the prodrug Metronidazole (Mtz) into a cytotoxic DNA cross-linking agent. To investigate the efficiency of this system, we targeted three different cell lineages in the heart, pancreas, and liver. Expression of the fusion protein Cyan Fluorescent Protein–NTR (CFP-NTR) under control of tissue-specific promoters allowed us to induce the death of cardiomyocytes, pancreatic β-cells, and hepatocytes at specific times. Moreover, we have observed that Mtz can be efficiently washed away and that, upon Mtz withdrawal, the profoundly affected tissue can quickly recover. These findings show that the NTR/Mtz system is effective for temporally and spatially controlled cell ablation in zebrafish, thereby constituting a most promising genetic tool to analyze tissue interactions as well as the mechanisms underlying regeneration. Developmental Dynamics 236:1025–1035, 2007. © 2007 Wiley-Liss, Inc.


Conditional targeted ablation is a powerful tool to study the role of a specific cell lineage or tissue in developmental or physiological processes. Moreover, the ability to temporally and spatially control tissue damage and genetically remove a specific cell population has great applications for regeneration studies.

A wide range of genetic cell ablation techniques has been developed and used in different model systems (reviewed in Lewandoski, 2001; McGuire et al., 2004). In zebrafish, however, to date, no genetic cell ablation tool has proven to be (1) spatially controllable and strictly confined to the target cell population, (2) temporally inducible, (3) germline transmissible, and (4) reversible. Zebrafish combines remarkable regenerative capacity (reviewed in Akimenko et al., 2003; Poss et al., 2003; Del Rio-Tsonis and Tsonis, 2003) with strong genetics, differentiating it from other regenerative model organisms such as urodele amphibians (Brockes and Kumar, 2005; Odelberg, 2005) and planarians (Sanchez Alvarado, 2006). The development of an inducible, targeted cell ablation methodology would, therefore, not only expand developmental studies of cell function and tissue interaction, but also optimize the use of zebrafish as a vertebrate model for studies of regeneration and stem cell regulation.

Although the well-established methods of physical surgery and laser-mediated ablation (Yang et al., 2004; Gahtan and Baier, 2004) are controllable in time and space, they are labor-intensive, time-consuming, and not as reproducible as a genetic approach. These methods are, therefore, limited when analysis of a large number of samples is required, such as during genetic or pharmacological screens. Attempts to develop stable genetic cell ablation tools using Diphtheria toxin A-chain (DTA; Han et al., 2000) have thus far been unsuccessful in zebrafish. Because of its high toxicity, minimal leakiness of the promoter used results in unintended cell death and the consequent failure to generate stable transgenic lines carrying DTA. DTA-mediated ablation has been successful in the zebrafish lens (Kurita et al., 2003) and exocrine pancreas (Wan et al., 2006), but only in transient transgenic embryos. To overcome the toxicity problem, another strategy to ablate cells used the prokaryotic parD system, which consists of expressing a toxin (kid) in the target cells and a prophylactic antidote (kis) in all other cells (Ruiz-Echevarria et al., 1991). Although this approach was successful in ablating primordial germ cells, these experiments were only performed in transiently kid/kis-expressing embryos (Slanchev et al., 2005). Inducibility would surmount the problems associated with unintended toxicity. The Tetracycline (Tet) -dependent transcriptional activation system is an inducible system that has been tested in zebrafish. However, weak expression in the absence of the inducer (leakiness) has been observed (Huang et al., 2005)—a common problem in Tet-regulated systems in many model organisms (Corbel and Rossi, 2002; Pluta et al., 2005).

The need for an effective targeted conditional ablation technique led us to test the ability of the Escherichia coli Nitroreductase (NTR)/Metronidazole (Mtz) [1-(2-hydroxyethyl)-2-methyl-5-nitroimidazole] system to genetically ablate cells in a specific and inducible manner in zebrafish. This system relies on the NTR-mediated conversion of a nontoxic prodrug (a nitroimidazole substrate such as Mtz) into a cytotoxic agent. The NTR enzyme is first reduced by nicotinamide adenine dinucleotide (NADH) or nicotinamide adenine dinucleotide phosphate (NADPH), and upon binding, Mtz is reduced by NTR and thereby converted into a potent DNA interstrand cross-linking agent, causing cell death (Lindmark and Muller, 1976; Anlezark et al., 1992; Edwards, 1993; Fig. 1A). Although the Thymidine kinase (Tk)/Ganciclovir cell ablation system relies on a similar principle, the converted toxic Ganciclovir metabolites act by inhibiting DNA polymerase, limiting it to proliferating cells (Springer and Niculescu-Duvaz, 2000). In contrast, the NTR/nitroimidazole system is cell-cycle independent and, therefore, applicable to any target cell population (Bridgewater et al., 1995).

Figure 1.

Tissue-specific Nitroreductase/Metronidazole (NTR/Mtz) cell ablation system. A: Mechanism of NTR/Mtz cell ablation: a cell expressing the bacterial NTR (NTR+; in cyan blue) when exposed to the prodrug Mtz (in green) converts the latter into a cytotoxic agent (in red), which causes DNA damage and death of the NTR+ cell. B,D: Constructs used to generate the Tg(cmlc2:CFP-NTR)s890 and Tg(l-fabp:CFP-NTR)s891 lines contain the cmlc2:CFP-NTR and l-fabp:CFP-NTR sequences flanked by tol2 sequences. C: The Tg(ins:CFP-NTR)s892 sequence is flanked by I-Sce sites (“i”). E–G: Brightfield combined with fluorescence imaging of 4, 3, and 6 days postfertilization (dpf) larvae, respectively, showing stable expression of Cyan Fluorescent Protein–NTR (CFP-NTR; reported by CFP fluorescence) in cardiomyocytes (autofluorescence in the yolk, E), Insulin-producing pancreatic β-cells (F) and hepatocytes (G).

To date, the most commonly used nitroimidazole substrate for NTR-mediated cell ablation has been CB1954 [5-(aziridin-1-yl)-2,4-dinitrobenzamide]. The CB1954 metabolite has, however, been shown to have a “bystander effect” due to its cell-permeability, thereby also affecting cells neighboring those expressing NTR (Bridgewater et al., 1997).

To investigate the potential of NTR to specifically ablate a defined subset of cells in zebrafish, we generated stable transgenic lines robustly expressing a fusion protein composed of Cyan Fluorescent Protein and NTR (CFP-NTR) in three different cell lineages: cardiomyocytes, pancreatic β-cells, and hepatocytes. Upon Mtz-treatment of the transgenic larvae, we observed rapid cell ablation restricted to the CFP-NTR expression domain, indicating that the genetic/pharmacological inducible NTR/Mtz system can be used for efficient cell ablation in zebrafish. Cell death of cardiomyocytes led to heart failure. Strikingly, the morphology, patterning, and contractility of the heart could be recovered and blood circulation reestablished upon drug withdrawal, showing that zebrafish larvae can restore functional cardiomyocytes. Additionally, we observed that, after CFP-NTR–mediated ablation of larval pancreatic β-cells, Insulin-expressing cells reappeared rapidly upon removal of Mtz from the medium.


Expression of CFP-NTR Under the Control of Three Different Tissue-Specific Promoters

To test the efficacy of the NTR/Mtz system for temporally and spatially controlled cell ablation in zebrafish, we generated stable transgenic lines in which a CFP-NTR fusion protein (CFP-NTR) is specifically expressed in three different cell types: cardiomyocytes, using the cardiac myosin light chain 2 (cmlc2) promoter (Huang et al., 2003; Fig. 1B); pancreatic β-cells, using the insulin promoter (Fig. 1C); and hepatocytes, using the liver-type fatty acid binding protein (l-fabp) promoter (Her et al., 2003; Fig. 1D). The generated transgenic lines stably express CFP specifically in the heart (cardiomyocytes; Fig. 1E), pancreatic β-cells (Fig. 1F), and liver (hepatocytes; Fig. 1G).

Conditional Ablation of Cardiomyocytes Results in Severe Cardiac Morphological and Functional Phenotypes

To test the effectiveness of the NTR/Mtz system in ablating cardiomyocytes, we exposed Tg(cmlc2:CFP-NTR)s890 48 hours postfertilization (hpf) embryos (sorted by CFP expression) to the prodrug Mtz. Twenty-four hours after incubation with 10 mM Mtz, all of the Tg(cmlc2:CFP-NTR)s890 larvae showed a severe cardiac phenotype: both the atrium and ventricle were collapsed and blood circulation had stopped (Supplementary Movies B[t1] and C[t1], which can be viewed at http://www.interscience., leading to blood pooling (Fig. 2B; arrow).

Figure 2.

Tissue-specific damage to the heart upon exposure of Tg(cmlc2:CFP-NTR)s890 larvae to Metronidazole (Mtz). A,B: Brightfield images of a control Tg(cmlc2:GFP) larva (A) and Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larva (B) exposed to 10 mM Mtz (in 0.2% dimethyl sulfoxide [DMSO]) at 48 hours postfertilization (hpf) for 24 hr. A′,B′: Fluorescence images of the larvae in A,B, showing Tg(cmlc2:GFP) expression (green). C–J: Confocal images of the heart taken after Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae were exposed to 0.2% DMSO alone (C,E,H), or 10 mM Mtz in 0.2% DMSO (D,F,G,I,J), from 58 to 76 hpf. C,D: Ventricles of DMSO (C) and Mtz-treated (D) Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae stained for F-actin (red), Tg(cmlc2:GFP) expression shown in green. E–G: Heart sections of DMSO (E) and Mtz-treated (F,G) Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae immunostained for activated Caspase-3 (blue) and stained with rhodamine phalloidin (red), Tg(cmlc2:GFP) in green. G: Same larva as in F at a higher magnification and thicker optical section. H–J: Heart sections of DMSO (H) and Mtz-treated (I,J) Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae processed for terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) detection (red), Tg(cmlc2:GFP) in green. Scale bars = 20 μm. Exposure of 48 hpf embryos to 10 mM Mtz for 24 hr did not affect control larvae (A,A′), but resulted in a severe functional cardiac phenotype in the CFP-NTR+ larvae (B,B′): collapsed atrium and ventricle, shown by Tg(cmlc2:GFP) expression (B′), and consequent failure in blood circulation with blood pooling (arrow in B). Confocal imaging of the CFP-NTR/Mtz–mediated damaged ventricle shows a significant loss of actin filaments as well as changes in the overall heart ventricle shape and dimensions (D), compared with ventricles of Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae exposed to DMSO only (C). Cell death was detected in hearts of Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae exposed to 10 mM Mtz—both by immunostaining for activated Caspase-3 (F,G) and TUNEL assay, in the ventricle (I) as well as in the atrium (J)—whereas in control Tg(cmlc2:CFP-NTR)s890 larvae treated with DMSO alone, no Caspase-3 expression (E) or TUNEL (H) was detected. TUNEL and Caspase-3 expression was detected exclusively in Tg(cmlc2:GFP)-expressing cells (F,G,I,J), indicating that cell death occurred specifically in cardiomyocytes and without affecting the neighboring CFP-NTR–negative cells.

For a better visualization of the heart morphology, we repeated these experiments in the Tg(cmlc2:GFP) background, which has higher fluorescence intensity in the cardiomyocytes. This transgene allowed us to better observe the effects on heart morphology (Fig. 2A′,B′) and analyze heart contractility. We found that, in Tg(cmlc2:GFP); Tg(cmlc2:CFP-NTR)s890 larvae treated with Mtz, the ventricle showed no contractility and the atrium poor or absent contractility (Supplementary Movies B′[t1] and C′[t1]). In all experiments, the ventricle was more sensitive to Mtz treatment and consistently the first cardiac chamber to stop contracting, and that within 24 hr. Control Tg(cmlc2:GFP) larvae (not carrying the cmlc2:CFP-NTR transgene) exposed to 10 mM Mtz showed no morphological defects (Fig. 2A,A′) or defects in heart contractility or blood circulation (Supplementary Movies A[t1] and A′[t1]). As an additional control, we exposed Tg(cmlc2:CFP-NTR)s890 larvae to 0.2% dimethyl sulfoxide (DMSO) from 48 to 72 hpf, as above, and observed no morphological or functional phenotypes (data not shown).

To gain greater insight into these morphological defects, we used confocal microscopy and observed that collapsed ventricles of Tg(cmlc2: CFP-NTR)s890; Tg(cmcl2:GFP) larvae treated with 10 mM Mtz from 58–76 hpf showed a dramatic change in their overall shape and dimension, as well as a significant loss of actin filaments (Fig. 2D) compared with the ventricles of Tg(cmlc2:CFP-NTR)s890; Tg(cmcl2:GFP) larvae exposed to DMSO alone (Fig. 2C). This loss of F-actin can explain the lack of contractility of the collapsed ventricles and consequent failure in blood circulation.

Because NTR activation of a nitroimidazole substrate (prodrug) has been shown to initiate apoptosis (Felmer and Clark, 2004), we investigated whether the cardiac phenotype was caused by apoptosis of CFP-NTR–expressing cardiomyocytes. We detected activated Caspase-3 exclusively in cardiomyocytes (green fluorescent protein–posiotive [GFP+] cells) of collapsed chambers in Tg(cmlc2:CFP-NTR)s890;Tg(cmlc2:GFP) larvae treated with Mtz from 58–76 hpf (Fig. 2F,G; Supplemental Figure S1B, which can be viewed at http://www.interscience., shows individual channels of activated Caspase [Supplemental Figure S1B′], F-actin [Supplemental Figure S1B″], and cmlc2:GFP [Supplemental Figure S1B′″] of image Fig. 2G, showing that the apoptotic cells are cardiomyocytes). In contrast, no activated Caspase-3 was detected in cardiomyocytes of Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae treated only with DMSO (Fig. 2E, Supplemental Figure S1A,A′″). We also detected terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) signal in cardiomyocytes of Mtz-treated CFP-NTR+ larvae (Fig. 2I,J) but not in DMSO-treated CFP-NTR+ (Fig. 2H) or Mtz-treated CFP-NTR larval hearts (data not shown). Furthermore, once cardiomyocytes started dying, GFP fluorescence and F-actin expression waned and membrane blebbing from the myocardial layer was observed (Fig. 2I,G). Exclusive overlap of these apoptosis indicators with GFP expression (Figs. 2F,G,I,J) indicates that cell death occurred specifically in cardiomyocytes. To further test for tissue specificity, we repeated the TUNEL assay and counterstained with TO-PRO, which marks all cell nuclei. Again, TUNEL signal was confined to the nuclei of cardiomyocytes, confirming that the neighboring cells were not undergoing apoptosis (data not shown). These results show that the NTR/Mtz system is effective in ablating cardiomyocytes and that CFP-NTR/Mtz–induced cell death occurred specifically in CFP-NTR–expressing cells.

Regeneration of Functional Myocardium After Severe Cardiomyocyte Damage

To investigate whether prodrug and CFP-NTR–converted (activated) Mtz could be cleared from the organism, we transferred Tg(cmlc2:CFP-NTR)s890 collapsed heart larvae to Mtz-free medium and monitored their cardiac phenotype as individuals in multiwell plates. Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae treated for 24 hr with Mtz (from 48 hpf) showing collapsed heart (Fig. 3C′) and blood pooling (Fig. 3C), as well as control larvae exposed to Mtz with no cardiac phenotype (Fig. 3A,A′), were transferred to Mtz-free medium. After 72 to 96 hr in Mtz-free medium, a majority (61%, n = 94) of the damaged-heart larvae showed wild-type body and heart morphology (Fig. 3D,D′) comparable to control larvae (Fig. 3B,B′). Moreover, the myocardial tissue that had been previously damaged and dysfunctional (Supplementary Movies B′[t1], C′[t1], B[t1], and C[t1]) now consisted of functional contracting tissue (Supplementary Movies B′[t2] and C′[t2]) and blood circulation was reestablished [Movies B(t2) and C(t2) - supplemental data], identical to age-matched nondamaged larvae [Movies A′(t2) and A(t2) - supplemental data]. In contrast, Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae with Mtz-induced cardiac phenotype that remained in Mtz showed morphological defects characteristic of lack of blood circulation, including large yolk edema, small eyes, enlarged eye-pockets, curved body (Fig. 3E), and nonfunctional heart of significantly reduced size (Fig. 3E′). Fifteen percent of the damaged-heart larvae placed in Mtz-free medium (n = 94) did recover cardiac chamber morphology but were not able to restore blood circulation (data not shown). We did not observe recovery, morphological or functional, in 24% of the larvae used in the recovery experiments (n = 94). We also observed that a later removal of Mtz from the medium resulted in a lower recovery rate, suggesting that damage that is too extensive or sustained is not reversible. These results not only indicate that it is possible to remove the insulting agent (Mtz) from the organism but also illustrate that, after severe injury, the nonfunctional larval heart can completely recover its cell mass, morphology, and function, within 3 to 4 days.

Figure 3.

Functional heart tissue recovery after Nitroreductase/Metronidazole (NTR/Mtz) -mediated cell ablation. A–D: Brightfield images of a control Tg(cmlc2:GFP) larva (A,B) and Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larva (C,D) exposed to 10 mM Mtz at 48 hours postfertilization (hpf), after 24 hr incubation in Mtz (t1; A,C) and after 96 hr washing-out in Mtz-free medium (t2; B,D). E: Brightfield image of Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larva incubated in 10 mM Mtz continuously from 48 to 168 hpf (t2). A′,B′,C′,D′,E′: Fluorescence images showing Tg(cmlc2:GFP) expression (green) in the larvae above (A′ corresponding to A, B′ to B, C′ to C, D′ to D, and E′ to E). Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae whose heart tissue was severely damaged as a result of exposure to Mtz for 24 hr (C, C′) were able to replace damaged tissue by functional tissue upon removal of the drug: after 96 hr in Mtz-free medium, the larvae showed a morphologically wild-type heart (D′) and no blood pooling (D). Tg(cmlc2:CFP-NTR)s890; Tg(cmlc2:GFP) larvae continuously exposed to Mtz for 120 hr showed enhanced cardiac and body phenotype defects: curved body, large yolk edema, small eyes, enlarged eye pockets (E), and collapsed, nonfunctional heart chambers of significantly reduced size (E′).

Ablation and Recovery of Pancreatic β-Cells

We also tested the efficacy of CFP-NTR and Mtz in ablating Insulin-producing pancreatic β-cells. Larvae containing the CFP-NTR transgene were identified by CFP fluorescence in the islet as early as 32 hpf and segregated from nontransgenic siblings. Tg(ins:CFP-NTR)s892 larvae were treated with control DMSO medium, and 0.5, 1, 5, 7.5, and 10 mM Mtz starting from 84 hpf. Only Mtz concentrations between 5 and 10 mM had consistent specific effects on pancreatic β-cells, and 5 mM was the concentration used for subsequent experiments. By 96 hpf (12 hr of treatment), neither control DMSO treatment, nor treatment of CFP-NTR larvae with Mtz had any observable effect on the pancreatic β-cells (Fig. 4A,A′, and data not shown). In contrast, incubation of transgenic larvae with Mtz resulted in reduced β-cell mass and activation of Caspase-3, an indicator of apoptotic induction (Fig. 4B,B′). In addition, Mtz-treated β-cells displayed a striking rounded cellular morphology (Fig. 4B).

Figure 4.

Ablation and recovery of pancreatic β-cells in zebrafish larvae. A,B: Confocal projections of 96 hours postfertilization (hpf) Tg(ins:CFP-NTR)s892 principle islets from larvae treated for 12 hr with dimethyl sulfoxide (DMSO) control (A) or Metronidazole (Mtz; B) and immunostained for CFP (cyan) and activated Caspase-3 (red). A: Caspase-3 staining is not observed in Cyan Fluorescent Protein-Nitroreductase–positive (CFP-NTR+) β-cells in DMSO-treated larvae. B: Strong staining for activated Caspase-3 and rounded β-cell morphology are seen in β-cells of Mtz-treated larvae (for clarity, A′ and B′ show only the Caspase-3 red channel fluorescence). C–H: Isolated pairs of Tg(ins:CFP-NTR)s892 larvae monitored during the course of DMSO control (C,E,G) or Mtz prodrug treatment (D,F,H) by combined brightfield/ fluorescence microscopy (scale bars = 250 μm). Inset images are confocal projections of the pancreatic islet of similarly treated sibling larvae, which have been immunostained for Insulin (green) and Glucagon (red; scale bars = 50 μm). C,D: Untreated 84 hpf Tg(ins:CFP-NTR)s892 larvae show strong CFP-NTR expression in the pancreatic islet (arrows). E,F: 110 hpf Tg(ins:CFP-NTR)s892 larvae, 26 hr after addition of 0.1% DMSO vehicle or 5 mM Mtz in 0.1% DMSO. E: Islet development appears unaffected after exposure to DMSO (arrows); a centralized “core” of Insulin+ β-cells is enveloped by a shell of Glucagon+ α-cells (inset). F: After exposure to Mtz, the majority of β-cells have been ablated (asterisks). The surrounding α-cells appear to collapse into the resulting void, forming a more condensed mass; remaining β-cells are clustered nearest the extrapancreatic duct (inset, arrowhead). G,H: 145 hpf Tg(ins:CFP-NTR)s892 larvae, 35 hr after washout of DMSO or Mtz. G: Pancreatic islet size and structure appears unaffected in DMSO (26 hr) /wash (35 hr) larvae (arrows and inset). H: Significant recovery of β-cell mass is evident in Mtz (26 hr) per wash (35 hr) larvae. Insulin+ β-cells repopulating the islet surround a centralized mass of Glucagon+ α-cells: an architecture inverse to the original one.

To investigate the regenerative capacity of the pancreatic islet after NTR/Mtz-mediated ablation, CFP-NTR+ larvae were maintained as pairs or single embryos in multiwell plates, enabling individual fish to be monitored over time. Before treatment, strong CFP-NTR expression was seen in the pancreatic islet at 84 hpf (Fig. 4C,D) and treatment with DMSO alone for 26 hr did not appear to affect the Tg(ins:CFP-NTR)s892 larvae (Fig. 4E). However, incubation of Tg(ins:CFP-NTR)s892 larvae with Mtz resulted in substantial reduction or complete ablation of fluorescence in Insulin+ cells (Fig. 4F). Confocal imaging revealed that this ablation was specific to β-cells, as adjacent Glucagon+ α-cells remained intact. However, when β-cell numbers were reduced, the arrangement of cells within the islet was affected: α-cells became more compacted, compared with their normal mantle-like distribution encapsulating β-cells (Fig. 4E,F, insets). Although some CFP-NTR fluorescence and Insulin protein persisted after 26 hr of Mtz treatment, these cells were likely in the process of apoptosing (see above). Moreover, any persistent β-cells detected after Mtz treatment were primarily found aggregated lateral to the islet, possibly reflecting a continual influx of newly formed Insulin+ cells from a source external to the principle islet, as previously described (Field et al., 2003).

To assess any regenerative capacity, the treated larvae were washed several times in fresh Mtz/DMSO-free medium and allowed to recover for 35 hr. The islet continued to develop normally in DMSO-treated larvae (Fig. 4G). In the β-cell ablated larvae, reconstitution of β-cell mass at 145 hpf was evident by restored CFP fluorescence and Insulin expression (Fig. 4H). Strikingly, confocal analysis showed that the newly added β-cells surrounded the aberrantly compacted core of α-cells, effectively inverting the relative positions of α- and β-cells in the islet (Fig. 4G,H inset). Further studies will be required to determine the mechanisms of islet recovery and the source of the newly formed β-cells.

Hepatocyte Damage Using the NTR/Mtz System

To investigate whether the NTR/Mtz system could be used to conditionally ablate hepatocytes in zebrafish larvae, we treated larvae stably expressing CFP-NTR under the control of the liver-specific l-fabp promoter with Mtz. Tg(l-fabp:CFP-NTR)s891 larvae and control larvae were treated with 10 mM Mtz, or DMSO alone, from 4 days postfertilization (dpf) to 7 dpf. By confocal imaging, we observed that hepatocytes of Tg(l-fabp:CFP-NTR)s891 larvae treated with Mtz had significantly enlarged nuclei (visualized by expression of the hepatic nuclear marker Prox1) and exhibited an abnormal cell morphology (visualized by F-actin staining; Fig. 5D). These phenotypes were not observed in control larvae: CFP-NTR DMSO-treated (Fig. 5A), CFP-NTR Mtz-treated (Fig. 5B), or CFP-NTR+ DMSO-treated (Fig. 5C), where hepatocytes maintained their size and shape. Additionally, we observed that hepatic cells that did not express CFP-NTR, such as bile duct cells, which sit adjacent to CFP-NTR–expressing hepatocytes, were not affected by Mtz treatment. That is, we did not observe any effect on the nuclear shape or Prox1 expression levels in the Alcam-positive cells, which at this stage form the intrahepatic biliary network (Fig. 5F). The single Prox1 channel of the confocal image shown in Figure 5F (Fig. 5F′) illustrates the enlarged nuclear size and seemingly reduced Prox1 expression level of CFP-NTR–expressing hepatocytes compared with the neighboring bile duct cells, and control CFP-NTR–nonexpressing hepatocytes treated with Mtz (Fig. 5E, E′). These results show that the NTR/Mtz system can have a specific effect on hepatocyte nuclei when Tg(l-fabp:CFP-NTR)s891 larvae are exposed to Mtz. To test whether the dysmorphic Mtz-treated hepatocytes were undergoing apoptosis, we performed a combined immunostaining/TUNEL analysis on transverse sections of Tg(l-fabp:CFP-NTR)s891 and wild-type larvae, Mtz-, or DMSO-treated, from 4 to 7 dpf. The sections were also immunostained for Prox1 (to mark all liver cells) and Alcam (to mark bile duct cells). Extensive cell death was detected specifically in liver tissue of Tg(l-fabp:CFP-NTR)s891 Mtz-treated larvae (Fig. 5H,I,I′), whereas no evidence of ectopic cell death could be detected in other cells within the same section, in control liver sections of wild-type larvae exposed to Mtz (Fig. 5G), or Tg(l-fabp:CFP-NTR)s891 DMSO-treated larvae (data not shown). We also observed that, after exposure to Mtz, CFP-NTR+ livers became spherical and showed a more dispersed distribution of cells (Fig. 5H) as compared with wild-type livers exposed to Mtz (Fig. 5G) or CFP-NTR+ livers treated with DMSO alone (data not shown).

Figure 5.

Hepatocyte damage mediated by the Nitroreductase/Metronidazole (NTR/Mtz) system. A–D: Confocal images of hepatic tissue of 7 days postfertilization (dpf) wild-type (A,B) and Tg(l-fabp:CFP-NTR)s891 (C,D) larvae exposed to dimethyl sulfoxide (DMSO) alone (A,C), or to10 mM Mtz (B,D) from 4 to 7 dpf, immunostained for Prox1 (blue) and F-actin (red). E,E′,F,F′: Confocal images of liver tissue of 5 dpf wild-type (E,E′) and Tg(l-fabp:CFP-NTR)s891 (F,F′) larvae exposed to 10 mM Mtz from 2 to 5 dpf and immunostained for Prox 1 (blue) and Alcam (red). E′,F′: Prox1 channel of images E and F, respectively. Scale bars = 10 μm. G–I: Confocal images of transverse sections of a 7 dpf wild-type control larva (G) and of a Tg(l-fabp:CFP-NTR)s891 larva exposed to 10 mM Mtz from 4 to 7 dpf (H; magnified liver section, I, I′) immunostained for Prox1 (green), Alcam (blue), and terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL, red). I′: Single channel confocal image of I showing TUNEL staining. A–D: Hepatocytes expressing CFP-NTR under the control of the l-fabp promoter, after treatment with Mtz from 4–7 dpf (D) show a different morphology from the respective controls: CFP-NTRDMSO (A), CFP-NTRMtz (B), and CFP-NTR+ DMSO (C). In the damaged livers, hepatocytes show an abnormal shape and a significant increase in the size of the nuclei, marked by Prox1 expression (F,F′). However, bile duct cells (surrounded by Alcam expression) do not appear affected (in their nuclear shape or intensity of Prox1 expression; F,F′) compared with the CFP-NTR control (E,E′). Immunostaining and TUNEL detection assay on transverse sections of Mtz-treated Tg(l-fabp:CFP-NTR)s891 larvae (H) vs. Mtz-treated CFP-NTR larvae (G) reveal specific cell death in the liver (H,I,I′), whereas no, or almost no, cell death could be detected in nonhepatic tissue or in control larvae (G).


In this study, we have shown that the NTR/Mtz system can be used to effectively ablate target cell populations in zebrafish larvae. The experiments performed here also demonstrate the usefulness of this technique in further developing zebrafish as a model to study the process of regeneration. To date, no genetic tool had been established that allows one to conditionally, both temporally and spatially, ablate a specific target cell population in zebrafish. While several genetic techniques have been directed toward specific cell ablation in zebrafish, they have all presented considerable limitations. Using the NTR/Mtz system, we have successfully targeted, in a conditional manner, three cell lineages in the heart, pancreas, and liver of zebrafish larvae.

We have shown that exposure of larvae expressing CFP-NTR under the myocardial promoter cmlc2 to the prodrug Mtz resulted in cell death of cardiomyocytes, without apparently affecting cells in the tissues adjacent to the myocardium. This result illustrates that this system can be used to ablate a cell lineage specifically, and further suggests that once the prodrug Mtz is activated by NTR, the cytotoxic derivatives are confined to the CFP-NTR–expressing cells. Our results showing NTR/Mtz–mediated pancreatic β-cell and hepatocyte damage confirm the specificity of this technique. These data are of particular importance because, although NTR has been used to ablate neurons (Isles et al., 2001), astrocytes (Cui et al., 2001), and adipocytes in mice (Felmer et al., 2002) and in multiple “gene suicide” cancer therapy studies (Searle et al., 2004), the prodrug used in those studies—CB1954—was shown to have a significant “bystander effect.” Once CB1954 is activated by NTR, its cytotoxic derivative diffuses into neighboring cells, causing nonspecific cell death (Bridgewater et al., 1997). Mtz has been previously used as an NTR substrate in studies on eradication and resistance of Helicobacter pylori (de Korwin, 2004) and ex vivo studies (for elimination of human cytotoxic T lymphocytes; Verdijk et al., 2004) and as a system for identifying transcriptionally responsive genes in cell lines (Medico et al., 2001). However, there are no previous reports of its use in in vivo cell lineage ablation. Here, we provide evidence that the NTR and Mtz system can be effectively used in vivo in a vertebrate organism, such as zebrafish.

In each tissue analyzed, not all CFP-NTR+ targeted cells were ablated after 18–24 hr of exposure to Mtz, as shown by the absence of TUNEL signal in some (cmlc2:GFP) cardiomyocytes and some targeted hepatocytes (Prox1+Alcam, or hepatocytes whose CFP-NTR expression was marked using a GFP antibody that recognizes CFP; data not shown) and by the detection of Insulin-expressing cells after Mtz-treatment. Several nonexclusive possibilities may account for this observation, including (1) a high turnover of the targeted cell population, with dying cells rapidly being replaced by new cells; and (2) differential sensitivity and response among the cells within the targeted tissue resulting in different rates of apoptosis.

The NTR/Mtz ablation technique is a robust binary system: for cell death to be induced, both the NTR enzyme and prodrug Mtz must be present. This concept was illustrated both by the lack of phenotype or ectopic cell death when CFP-NTR was expressed in the absence of Mtz, and when larvae not expressing CFP-NTR were exposed to Mtz. As opposed to the DTA technology, which requires very tight expression control, this combined genetic and pharmacological system is a major advantage, as it allows germline transmission and generation of stable NTR-expressing transgenic lines. The availability of stable transgenic lines will, therefore, allow the design and execution of reproducible cell ablation assays. Moreover, such a genetic tool will be of particular utility when it is necessary to ablate cells in a large number of animals, as it will provide not only reproducibility, but also simplicity and synchronicity to the ablation assay. This method stands in contrast to more laborious and time-consuming ablation methods, such as individual laser ablation or tissue removal by surgery, which are somewhat impractical for large-scale analyses.

Another strength of this technique is its versatility, that is, its applicability in a wide variety of cellular targets. In this study, we have shown that the NTR/Mtz system was effective in ablating cells of three different lineages. This result suggests that this tool can be applied to any cell population, and its applicability should be limited only by the availability of cell type-specific promoters. Several other approaches have been successfully used for ablation of certain cell types: for example, small molecule-induced ablation of melanocytes (Yang and Johnson, 2006) or using specific hepatotoxic chemicals for liver damage and regeneration studies in mouse (Fausto et al., 2006). However, their applicability is, by nature, limited to a certain tissue or cell population. Therefore, it will be advantageous to have available a more universal conditional ablation technique such as the one presented here.

In addition to the flexibility of promoter selection, and thus the target cell lineage to be ablated, the NTR/Mtz system is versatile with regard to the timing of cell death induction, as it is entirely dependent upon the addition of the prodrug. Thus, the combination of genetically controlled NTR expression with the pharmacological inducibility of the system makes it a highly versatile system with a broad scope of applications.

Addition of Mtz to Tg(cmlc2:CFP-NTR)s890 48 hpf embryos induced the death of cardiomyocytes and resulted, after 18–24 hr, in a severe functional cardiac phenotype: collapsed atrium and ventricle, failure in heart contractility, and consequent absence of blood circulation. A striking result of this study was the functional recovery of the damaged heart once Mtz was removed from the medium: within 72–96 hr in Mtz-free medium, both atrium and ventricle recovered their morphology, patterning, and contractility, and blood circulation had been re-established. The adult zebrafish heart has been shown to have the capability to regenerate surgically removed ventricular tissues (Poss et al., 2002). We show that the larval heart is also able to regenerate and that at least part of the damaged tissue is replaced by functional contractile tissue. Furthermore, we show that the pancreatic β-cells are specifically ablated after Mtz treatment for 12–24 hr and that, after removal of Mtz from the medium, the β-cells mass dramatically recovered. These results indicate that Mtz can be washed away from the organism. The reversibility of the NTR/Mtz system adds to its advantages and makes it a powerful tool for regeneration studies.

The advantages of the NTR/Mtz ablation system—versatility, specificity, inducibility, genetic transmissibility, and reversibility—combined with the strength of forward genetics in zebrafish and its remarkable regenerative capabilities, open new possibilities for designing large-scale forward genetic or chemical screens aimed at studying tissue- and cell-specific regeneration of various organs in a vertebrate model system. Zebrafish larvae have been shown to be a useful model for studying regeneration of melanocytes (Yang et al., 2004; Yang and Johnson, 2006), hair cells (Harris et al., 2003), and fin, which was shown to regenerate by a mechanism similar to that at play during adult regeneration (Kawakami et al., 2004a). Our data suggest that zebrafish larvae can be used to study heart as well as pancreatic β-cell regeneration.

The transparency of zebrafish embryos/larvae at the stages used in this study combined with noninvasive fluorescent reporters makes it possible to clearly monitor damage as well as regeneration of internal and, therefore, less accessible tissue, such as the heart and pancreatic β-cells. Furthermore, the zebrafish embryo/larva as a model system provides a great advantage for the study of heart regeneration, as the embryo/larva receives enough oxygen through diffusion during the first week of development, making it possible to interfere with its cardiac function without affecting its survival (Stainier, 2001). Finally, the specificity of the NTR/Mtz system and absence of a “bystander effect” should encourage its use in “gene suicide” cancer therapies where specificity is preferred, as well as its application in cell or tissue interactions and regeneration studies in other model organisms.


Nitroreductase Constructs and Zebrafish Lines

Zebrafish embryos, larvae, and adult fish were raised under standard laboratory conditions at 28°C (Westerfield, 2000). We used the transgenic line Tg(cmlc2:EGFP) (Huang et al., 2003), and we generated the Tg(cmlc2:CFP-NTR)s890, Tg(l-fabp:CFP-NTR)s891, and Tg(ins:CFP-NTR)s892 transgenic lines. Nitroreductase was amplified directly from Escherichia coli using the oligos 5′-ATG CTC GAG CCA TGG ATA TCA TTT CTG TCG CCT TA and 5′-GGG GAT CCG ATC GAT CTC AAT ACC CGC TAA ATA, to introduce XhoI and BamHI sites, respectively. The resultant (∼700-bp) product was directionally inserted into the pECFP vector (Clontech, Palo Alto, CA), following XhoI/BamHI double-digest, to create the pECFP-Nitro plasmid. To express the fusion protein CFP-NTR in cardiomyocytes, we generated the Tg(cmlc2:CFP-NTR)s890 construct by excising the 800-bp cmlc2 promoter from pBISceI-cmlc2 plasmid (gift of Ian Scott) with ApaI and XbaI, and subsequently inserting it into the tol2_CFP_Nitro plasmid (gift of Stephen L. Johnson), previously digested with ApaI and NheI. For expression of CFP-NTR in hepatocytes, the Tg(l-fabp-CFP-NTR) construct was generated: a 2.8-kb fragment of the l-fabp promoter was PCR-amplified from the pLF2.8-EGFP plasmid (Her et al., 2003) with the oligos 5′-GGG CCC TGA GCA TCA GAA TGG GGA AGG and 5′-GCT AGC GCT CAA CAC AAA GTG AAG GTC AGC, digested with ApaI and NheI, and then inserted into the tol2_CFP_Nitro plasmid, previously digested with ApaI and NheI. For genome integration and generation of stable transgenic lines, we coinjected the Tg(cmlc2:CFP-NTR)s890 or Tg(l-fabp:CFP-NTR)s891 plasmids together with Tol2 Transposase RNA (Kawakami et al., 2004b) into one-cell stage embryos. To express the CFP-NTR fusion protein in pancreatic β-cells, a 1.2-kb SacI-BamHI fragment of the zebrafish insulin promoter was first relocated from an insulin:GFP plasmid (gift of W. Driever) to a version of pBluescriptII SK+ in which the MCS was flanked by I-SceI restriction sites. Subsequently, to create pBiSKi (ins:CFP-NTR), the CFP-NTR coding sequence was amplified with the oligos 5′-GGC GAA TTC ATG GTG AGC AAG GGC GAG GAG and 5′-GCA TGT CGA CTG CCG ATT TCG GCC TAT TGG, cut with BamHI and SalI, and inserted into pBiSKi-ins downstream of the insulin promoter. This plasmid was injected into one-cell stage embryos together with I-SceI meganuclease as described (Grabher et al., 2004). Tg(cmlc2:CFP- NTR)s890, Tg(l-fabp:CFP-NTR)s891, and Tg(ins:CFP-NTR)s892 adult carriers were identified by screening their progeny for CFP fluorescence.

Mtz Preparation and Treatment

Mtz (Sigma catalog no. M1547) was dissolved in 0.1% or 0.2% DMSO in standard zebrafish embryo medium with vigorous shaking and was protected from ambient light. When treating wild-type larvae starting at 72 hpf we first observed nonspecific death of larvae at 24 hr after continuous exposure to 20 mM Mtz (19%, n = 37), and 48 hr after exposure to 15 mM Mtz (40%, n = 38). When treating wild-type embryos starting at 48 hpf we first observed death of larvae at 48 hr, and 60 hr, after exposure to 20 mM (57%, n = 29) and 15 mM Mtz (13%, n = 39), respectively. In contrast, we observed that, when treating wild-type zebrafish starting at 48 or 72 hpf, no larvae died after at least 60 hr of continuous exposure to 5 (n = 34) or 10 mM (n = 35) Mtz. Therefore, the Mtz concentrations used in the experiments in the present study were 5 or 10 mM.

To ablate cardiomyocytes or hepatocytes, Tg(cmcl2:CFP-NTR) or Tg(l-fabp:CFP-NTR)s891 zebrafish embryos/larvae were treated with freshly prepared 10 mM Mtz 0.2% DMSO at 28°C in the dark. As controls, Tg(cmcl2:CFP-NTR) or Tg(l-fabp:CFP-NTR)s891 embryos/larvae were incubated in 0.2% DMSO in embryo medium and wild-type CFP-NTR embryos/larvae were incubated with 10 mM Mtz. To ablate pancreatic β-cells, Tg(ins:CFP-NTR)s892 embryos/larvae were treated with freshly prepared 5 mM Mtz, 0.1% DMSO at 28°C in the dark, although similar results were obtained with 10 mM Mtz in 0.2% DMSO. As controls, Tg(ins:CFP-NTR)s892 embryos/larvae were treated with DMSO alone, and wild-type CFP-NTR embryos/larvae were treated with Mtz. For recovery experiments, the Mtz-containing medium was replaced with several changes of fresh embryo medium, and embryos/larvae were returned to 28°C.

Sectioning, Immunohistochemistry, and Microscopy

Larvae were fixed overnight at 4°C with 3% formaldehyde and then immunostained, in whole-mounts or after sectioning. For sectioning, larvae were embedded in low-melting agarose and cut into 200-μm sections with a Vibratome. Antibody staining was carried out in PBST (0.1% Triton X-100, 4% bovine serum albumin [BSA], 0.02% NaN3 in phosphate buffered saline [PBS] pH 7.3). We used the following antibodies: polyclonal antibody against Prox1 (rabbit, 1:1,000; Chemicon), Zn8 monoclonal antibody against Alcam (Dm-Grasp; mouse, 1:10, Developmental Studies Hybridoma Bank, University of Iowa), polyclonal antibody against activated Caspase-3 (rabbit, 1:100, Chemicon), monoclonal antibody against Glucagon (mouse, 1:200, Sigma), polyclonal antibody against Insulin (guinea pig, 1:100, Biomeda), polyclonal antibody against GFP (chicken, 1:1,000, Aves Labs), and fluorescently conjugated secondary antibodies from Molecular Probes. For F-actin or nuclear staining, whole larvae or sections were stained with rhodamine phalloidin (1:100, Molecular probes) or TO-PRO3 (1:5,000, Molecular Probes), respectively. Images were acquired with a Zeiss Pascal confocal or Zeiss Lumar microscope and were prepared for publication using Adobe Photoshop and Illustrator.

TUNEL Cell Death Assay

The TUNEL cell death assay was performed using the In Situ Cell Death Detection Kit, TMR red (Roche catalog no. 12156792910). After fixation of zebrafish larvae in 3% formaldehyde and sectioning, heart and liver transverse sections were preincubated in PBST, and then labeled with the TUNEL kit for 1 hr at 37°C. Sections were washed with PBT (0.1% Triton X-100 in PBS pH 7.3) and visualized by confocal imaging. For coimmunostaining with Prox1 and Alcam, sections were first incubated with primary antibodies, then with TUNEL solutions, and finally with secondary antibodies.


We thank Stephen L. Johnson (Washington University) for his suggestion regarding the use of the NTR/Mtz system in zebrafish; Lara Gnügge and Wolfgang Driever for the insulin:GFP construct; A. Ayala, S. Waldron, and N. Zvenigorodsky for expert help maintaining the fish; members of the Stainier lab for critical comments on the manuscript; and Mike Parsons and Steve Leach for coordinating publications. Construction of the CFP-NTR plasmid was supported by an NIH grant to Rachel O.L. Wong and NRSA postdoctoral fellowship (NIH/NEI) to J.S.M. J.S.M. and E.H.S. are cofounders and shareholders of Luminomics, Inc. R.M.A. was supported by a postdoctoral fellowship from the JDRF. D.Y.R.S. was funded by grants from the NIH (NHLBI and NIDDK) and the Packard Foundation.