Conditional Gene Expression and Lineage Tracing of tuba1a Expressing Cells during Zebrafish Development and Retina Regeneration

Abstract

The tuba1a gene encodes a neural-specific alpha-tubulin isoform whose expression is restricted to the developing and regenerating nervous system. Using zebrafish as a model system for studying CNS regeneration we recently showed that retinal injury induces tuba1a gene expression in Müller glia that reentered the cell cycle. However, due to the transient nature of tuba1a gene expression during development and regeneration, it was not possible to trace the lineage of the tuba1a-expressing cells with a reporter directly under the control of the tuba1a promoter. To overcome this limitation, we generated tuba1a:CreER^T2 and β-actin2:loxP-mCherrry-loxP-GFP double transgenic fish that allowed us to conditionally and permanently label tuba1a-expressing cells via ligand-induced recombination. During development, recombination revealed transient tuba1a expression in not only neural progenitors, but also cells that contribute to skeletal muscle, heart and intestine. In the adult, recombination revealed tuba1a expression in brain, olfactory neurons and sensory cells of the lateral line, but not in the retina. Following retinal injury, recombination showed tuba1a expression in Müller glia that had reentered the cell cycle and lineage tracing indicated these cells are responsible for regenerating retinal neurons and glia. These results suggest that tuba1a-expressing progenitors contribute to multiple cell lineages during development and that tuba1a-expressing Müller glia are retinal progenitors in the adult.

Introduction

Unlike mammals, adult teleost fish like zebrafish, are able to regenerate a damaged central nervous system and therefore provide a good model for investigating mechanisms underlying neuronal repair. The retina is a readily accessible portion of the central nervous system that has served as a model for studying both axonal and neuronal regeneration in fish (Hitchcock et al., 2004; Becker and Becker, 2007, 2008). The robust regenerative power of the teleost retina is demonstrated by the observation that chemically-induced whole retina destruction results in retinal regeneration that is accompanied by restoration of visual function (Sherpa et al., 2007). The source of cells that contribute to retinal regeneration appear to be resident in the inner nuclear layer where clusters or columns of proliferating cells are observed (Raymond and Hitchock, 2000; Vihtelic and Hyde, 2000; Wu et al., 2001; Faillace et al., 2002). The observation that inner nuclear layer cell proliferation precedes retina regeneration suggested that retinal stem cells residing in the inner nuclear layer are the predominant source of progenitors for regeneration. Interestingly, Müller glia, whose cell bodies reside in the inner nuclear layer, were observed to reenter the cell cycle following retinal damage (Braisted et al., 1994; Cameron, 2000; Wu et al., 2001). These observations, combined with reports that Müller glia in postnatal chicks and rodents have a limited capacity to regenerate neurons following retinal injury (Fischer and Reh, 2001; Ooto et al., 2004), implicated Müller glia as a source of retinal progenitors that participate in repair of damaged retinas.

Support for Müller glia as a source of retinal progenitors comes from studies using transgenic fish where the fate of GFP-expressing Müller glia can be followed for short periods of time following retinal injury (Fausett and Goldman, 2006; Bernardos et al., 2007; Fimbel et al., 2007; Thummel et al., 2008). Using transgenic fish harboring a 1016bp fragment of the tuba1a promoter driving GFP expression (1016 tuba1a:GFP) we found that tuba1a-expressing Müller glia respond to retinal injury by dedifferentiating and proliferating (Fausett and Goldman, 2006). However, because the tuba1a promoter was only transiently expressed in these dedifferentiated Müller glia it was not possible to follow their fate over long periods of time and confirm they were stably integrated into the retinal architecture. To follow the fate of these tuba1a-expressing progenitors in the adult retina we developed a conditional expression system that allowed for permanent labeling of tuba1a-expressing Müller glia. Based on previous success in mice, we chose to use CreER^T2/LoxP recombination system for this purpose (Branda and Dymecki, 2004).

Cre-mediated recombination offers many advantages for studying gene function by stimulating targeted deletion, insertion, inversion and exchange of chromosomal DNA (Branda and Dymecki, 2004). Recently the Cre/LoxP system has been used for conditional gene expression in developing zebrafish where either a tissue specific or a ubiquitously expressed heat shock-regulated promoter was used to drive Cre recombinase expression (Feng et al., 2007; Le et al., 2007; Liu et al., 2008; Wang et al., 2008; Hans et al., 2009; Collins et al., 2010). However, these promoters have limitations since they either lack temporal control (tissue-specific promoters) or may exhibit leaky expression (heat shock promoters). In addition, stressing fish by a heat shock may have unintended consequences on cell function. In the mouse these limitations have been surmounted by taking advantage of a ligand-dependent chimeric Cre recombinase where Cre is fused to the mutant ligand-binding domain of the human estrogen receptor (CreER^T and CreER^T2) (Feil et al., 1996; Feil et al., 1997; Danielian et al., 1998). These chimeric Cre recombinases are efficiently activated by the synthetic estrogen receptor ligand 4-hydroxytamoxifen (4-OHT) but are insensitive to endogenous 17β-estradiol (Indra et al., 1999).

Recently transgenic zebrafish were created harboring the CreER^T2 transgene under control of different promoters (Boniface et al., 2009; Hans et al., 2009). One study, using fish embryos that harbor the pax2a:CreER^T2 transgene and a recombination reporter driven by the EF1a promoter, found ligand-dependent CreER^T2 activation and recombination in developing embryos (Hans et al., 2009). In contrast, another study, using the her4.1 promoter to drive CreER^T2 expression, reported ligand-independent CreER^T2 activation that could be prevented by appending an additional ER ligand binding domain to the CreER^T2 fusion (Boniface et al., 2009). These studies were restricted to the evaluation of conditional gene expression in early developing embryos by bathing fish embryos in water containing 4-OHT. Although these studies suggest the CreER^T2/LoxP system will be useful for conditional gene expression during development, its suitability for conditional gene expression in adults and its use for lineage tracing in developing and adult animals remained untested.

Motivated by the need for a conditional gene expression system that would allow gene recombination at any stage of development, including adults, and that was amenable for lineage tracing of Müller glia-derived progenitors in the injured retina, we developed the following transgenic fish: 1) 1016 tuba1a:CreER^T2, where a 1016bp fragment of the tuba1a promoter directs CreER^T2 expression to the developing and regenerating CNS; and 2) β-actin2:LCLG, where the ubiquitously expressed β-actin2 promoter drives mCherry (C) expression that is flanked by loxP (L) sites and followed by an out of frame EGFP (G) sequence; and 3) 1016 tuba1a:CreER^T2;β-actin2:LCLG double transgenic fish where the latter transgene serves as a recombination reporter and allows one to permanently label cells that either constitutively or transiently express CreER^T2 driven by the 1016 tuba1a promoter. Using these fish, we show that transgenic lines expressing CreER^T2 at low levels do not exhibit basal ligand-independent CreER^T2 activity. These low expressing lines allowed us to map the fate of cells expressing the 1016 tuba1a promoter during development and in the adult injured retina. We found that this recombination system revealed very low and transient 1016 tuba1a promoter activity that could not be observed using traditional 1016 tuba1a:GFP transgenes. This improved sensitivity allowed us to identify descendents of tuba1a expressing cells early in development that include neural and non-neural progeny. In addition we show that this recombination system is suitable for conditional gene expression, which allows one to perform lineage analysis and assay the function of specific genes at any stage of zebrafish development. Using this conditional gene expression system, we mapped the fate of tuba1a expressing Müller glia in the injured retina and found they regenerate new retinal neurons and glia.

Materials and Methods

Zebrafish husbandry

Zebrafish were obtained from our breeding colony and maintained at 28 °C with a 10/14h light/dark cycle. Our fish originated from a local pet store. Zebrafish were treated in accordance with the guidelines of the University Committee on Use and Care of Animals at the University of Michigan.

Expression vectors and transgenic fish

The 1016 tuba1a:CreER^T2 expression vector (Supplementary Fig. 1) harbors 1016bp of 5’ flanking DNA from the goldfish tuba1a gene followed by exon 1 and intron 1 (Heiber et al., 1998), fused in-frame to the CreER^T2 sequence (Feil et al., 1997) and followed by an SV40 polyA signal sequence. The 1016 tuba1a promoter is active throughout the developing nervous system and in the adult retina this promoter is specifically activated in Müller glia-derived retinal progenitors following injury (Fausett and Goldman, 2006). Just downstream of this expression cassette we inserted a second expression cassette harboring the same sequences described above except the tuba1a 5’ flanking DNA was truncated to 906bp and CFP was inserted into the non-coding portion of exon 1, which was followed by intron 1 of the goldfish tuba1a gene. The reason we chose a shorter fragment of the tuba1a promoter over a heterologous promoter is that: 1) we did not want to risk influencing 1016 tuba1a promoter activity by novel regulatory elements in this second promoter; and 2) we already knew that the shorter tuba1a promoter fragment was poorly expressed during development and undetectable in the adult nervous system and injured retina (Fausett et al., 2008). The recombination reporter β-actin2:LCLG expression vector (Supplementary Fig. 1) harbors 3851bp of 5’ flanking DNA from the zebrafish β-actin2 promoter followed by exon 1 and intron 1 (Hagashijima et al., 1997) fused in frame with a LoxP (Branda and Dymecki, 2004) flanked mCherry sequence that is followed by an out of frame EGFP sequence and a SV40 polyA signal sequence. The plasmid backbone harboring these sequences was the Tol2 vector pT2AL200R150G (Urasaki et al., 2006). The tuba1a promoter directs gene expression to the developing and regenerating CNS (Hieber et al., 1998; Goldman et al., 2001; Senut et al, 2004).

Transgenic fish were produced by Tol2 transposase-mediated integration of the expression cassette as described in Urasaki et al., (2006). Injected embryos were analyzed for CFP or mCherry expression at 2-5 days post fertilization (dpf) using a Leica MZFLIII fluorescent stereo microscope and fish expressing the reporter gene were grown to adulthood. F0 fish were bred in groups of 6-8 fish and progeny were examined for reporter gene expression. Groups with fluorescent progeny were then bred individually with wild-type fish to identify founders with germ-line integration of the transgene. Progeny were raised to adults and bred to wild-type fish to perpetuate heterozygous lines. Multiple lines of fish harboring either the 1016 tuba1a:CreER^T2 or the β-actin2:LCLG transgenes were generated. Representative weak and strong expressing lines were saved and maintained for at least 4 generations without any noticeable change in transgene expression. 1016 tuba1a:CreER^T2 and β-actin2:LCLG fish were interbred and double transgenic fish were identified by CFP and mCherry fluorescence.

The 1016 tuba1a:GFP transgenic line (formerly referred to as 1016α1TIpEGFP and 1016α1T) harbors 1016bp of 5’ flanking DNA from the goldfish tuba1a gene followed by exon 1 and intron 1 fused in-frame to the GFP sequence and followed by an SV40 polyA signal sequence (Heiber et al., 1998; Fausett and Goldman, 2006).

Cell culture transfection

HEK 293 cells were cultured in DMEM with 10% fetal bovine serum and maintained at 5% CO₂ at 37 °C. Cells were transfected with Lipofectamine 2000 (Invitrogen) using 200 ng of recombination reporter and anywhere from 0.1-10 ng of tuba1a:CreER^T2 expression vector per 0.5 ml of culture media. Twenty-four hours after transfection 4-hydroxytamoxifen (4-OHT) was added to a final concentration of 1μM. Control cells received ethanol which is the vehicle used to dissolve 4-OHT. Cultured cells were observed 48h post 4-OHT addition using an inverted DMIL fluorescent microscope (Leica).

Retinal Injury

Retinas were injured as previously described (Senut et al., 2004; Fausett and Goldman, 2006; Fausett et al., 2008). Briefly, fish were anesthetized and under microscopic visualization, the right eye was gently pulled from its socket and poked 4 times; once in each quadrant through the sclera with a 30g needle. The needle was inserted to the length of the bevel (~5mm). The left eye served as an unoperated control.

4- hydroxytamoxifen (4-OHT) treatment and bromodeoxyuridine (BrdU) labeling

To activate CreER^T2, fish were immersed for various lengths of time, up to 3 days, in a small volume of fish water containing 1μM 4-OHT. Following 4-OHT immersion, fish were rinsed extensively in fish water prior to being returned to their tanks. During extended periods of 4-OHT treatment, fish were removed from 4-OHT containing solution and allowed to feed for a few hours before returning to fish water with 4-OHT. In some experiments, adult fish received 4-OHT simultaneously with retinal injury by injecting ~1 μl of a 50μM solution of 4-OHT with the same needle used to injure the retina. To identify dividing cells, fish were either immersed in 10mM BrdU (Sigma) for various amounts of time or anesthetized and injected intraperitoneally (IP) with 2-10 μl of a 25 mg/ml BrdU stock solution.

PCR analysis of reporter recombination and tuba1a:CreER^T2 splicing

Double transgenic embryos at 24 hours post fertilization (hpf) or adult fish whose retinas were injured 20h earlier, were either immersed in 1μM 4-OHT or vehicle and harvested at various times for analysis of recombination by RT-PCR. Embryos or adult injured retinas were lysed in TRIzol (Invitrogen) for RNA isolation. RNA was reverse transcribed using oligo(dT) primer and SuperScript II (Invitrogen) and 1/20^th of the resulting cDNA mixture served as template for PCR reactions. To assay recombination, we designed primers flanking the mCherry sequence: therefore the forward primer (TGCGGAATATCATCTGCTTG) targeted exon 1 of the β-actin2 gene and the reverse primer (GTTGTGGCGGATCTTGAAGT) targeted the EGFP sequence. We also assayed whether splicing of the1016 tuba1a:CreER^T2 derived transcript was accurate using primers that flanked the tuba1a intron; therefore, a forward primer that targeted exon 1 of the tuba1a gene (CTTACATCGATCCCTTAGTTGTCG) and a reverse primer targeting CreER^T2 (ATGTTTAGCTGGCCCAATG). Cycling conditions were 94 °C for 5 min followed by quick cooling on ice and the addition of Taq DNA polymerase; then cycling as follows: 15 seconds at 94 °C, 30 seconds at 62 °C and 1 min at 68 °C; a final extension at 68 °C for 5 min. We used 25 cycles for embryos and 35 cycles for adult retina.

In situ hybridization

In situ hybridization was carried out using antisense digoxigenin-labeled probes as previously described (Gulati-Leekha and Goldman, 2006). Briefly, fixed embryos were dechorionated and stored in methanol at -20 °C. Embryos were cleared in methanol-xylene and rehydrated in gradients of methanol (90-50%). Embryos were digested with Proteinase K for 5 min at 37 °C. Embryos were prehybridized for 2 hours at 56 °C and then hybridized with digoxigenin-labeled CreER^T2 (800bp coding sequence) or tuba1a (870bp 3’ UTR and exon4) antisense probe overnight at 56 °C. Posthybridization procedures were performed as previously described (Gulati-Leekha and Goldman, 2006). The anti-digoxigenin antibody conjugated to alkaline phosphatase was used at a 1:5000 dilution.

Tissue preparation and immunohistochemistry

For analysis of recombination in live embryos and adults we anesthetized the fish in 0.02% tricaine methane sulfonate (Sigma). Embryos were placed in 3% methylcellulose whereas adults were placed on a support and fluorescence was observed with a fluorescent stereo microscope (Leica MZFLIII) with attached digital camera for capturing images. For visualizing retinal sections, fish were overdosed with tricaine methane sulfonate and eyes from adult fish were enucleated, followed by removal of the lens and fixed by immersion in fresh 1% paraformaldehyde in 0.1M phosphate buffer, pH 7.4 for about 16 hours. After fixation, samples were cryoprotected in phosphate-buffered 20% sucrose before embedding with Tissue-Tek O.C.T. compound (Sakura, Finetek). Embedded samples were kept frozen at -80 °C until sectioned to 8 microns on a CM3050S cryostat (Leica). Sections were collected on Superfrost/Plus slides (Fisher Scientific), dried overnight at room temperature and stored at -80 °C. Immunohistochemistry was performed as previously described (Senut et al., 2004; Fausett and Goldman, 2006). For BrdU staining, sections were pretreated with 2N HCl for 20 min at 37 °C. Immunostained slides were washed in PBS and immersed in water containing 10 ng/ml of DAPI (Sigma) to visualize nuclei. Slides were rinsed with water and allowed to dry in the dark prior to coverslipping with 2.5% PVA (PVA-polyvinyl alcohol)/DABCO (1,4 diazabicyclo [2.2.2]octane). Slides were examined with a Zeiss Axiophot fluorescent microscope equipped with a digital camera or an Olympus FluoView FV1000 confocal imaging system.

Antibody characterization

Please see Table 1 for list of all antibodies used in this study. All antibodies have been previously characterized in zebrafish as described below and gave identical staining patterns in the current study.

Table 1.

Primary Antibodies

Antibody	Immunogen	Stains	Dilution	Host species	Manufacturer
BrdU monoclonal	BrdU	Proliferating cells	1:400	Rat	Abcam (ab6326)
zpr-1 monoclonal	Whole formaldehyde-fixed retinal cells	Red/green cones	1:250	Mouse	Zebrafish International Resource Center (zpr-1)
GFP polyclonal	Green fluorescent protein from A. victoria	Green fluorescent protein	1:1000	Rabbit	Invitrogen (A-6455)
HuC/D monoclonal	Human HuD peptide, QAQRFRLDNLLN	Amacrine and ganglion cells	1:500	Mouse	Invitrogen (A-21272)
Glutamine synthetase monoclonal	Sheep brain glutamine synthetase	Müller glia	1:500	Mouse	Chemicon/Millipore (MAB302)
PKC-β1 polyclonal	C-terminus peptide fragment of human protein kinase C β1 EFAGFSYTNPEFVINV	Bipolar cells	1:500	Goat	Santa Cruz Biotechnology (sc 209 G)
zn5 monoclonal	Crude extract of 1-5 day old zebrafish	Retinal ganglion cells	1:500	Mouse	Zebrafish International Resource Center (zn-5)

The BrdU antibody has been used to detect BrdU-labeled proliferating cells and does not detect any cells in animals that have not been treated with BrdU (Adolf et al., 2006; Goldman and Fausett, 2006; Grandel et al., 2006; Pellegrini et al., 2007).

The zpr1 antibody, formerly known as Fret43, identifies the red/green double cone pair in the outer nuclear layer (ONL) of the zebrafish retina as determined by anatomical location (ONL) and morphological features that include labeling of the outer segment, axon and synaptic terminal, and outlining of the inner segment (Larison and Bremiller,1990; Yazulla and Studholme, 2001); an identical staining pattern has been reported when this antibody was used to detect double cones in the adult zebrafish retina (Yazulla and Studholme, 2001; Fausett and Goldman, 2006; Raymond et al., 2006; Bernardos et al., 2007).

The GFP antibody only stains tissue from transgenic fish expressing the GFP transgene and no signal is detected from wild type fish that do not harbor the transgene (Fausett and Goldman, 2006; Fausett et al., 2008).

The HuC/D antibody identifies amacrine and ganglion cells in the mature rat retina as determined by anatomical localization, morphology, co-staining for the ganglion cell marker GAP-43 and lack of co-staining with bipolar-specific markers (Ekstrom and Johansson, 2003); an identical staining pattern has been reported when this antibody was used to detect amacrine and ganglion cells in the adult zebrafish retina where amacrine cells are concentrated in the proximal region of the inner nuclear layer (INL) and extend axons into the inner plexiform layer, while ganglion cells, with large round cell bodies, are concentrated in the ganglion cell layer (GCL) (Fausett and Goldman, 2006; Bernardos et al., 2007; Fimbel et al., 2007; Kassen et al., 2007).

The glutamine synthetase (GS) antibody recognizes a single 45-kDa protein in adult rat retinal tissue by Western blot analysis consistent with the size of glutamine synthetase and stains cells with morphological features uniquely characteristic of Müller glia, i.e. cell bodies in the INL along with processes that extend into the GCL and ONL (Chang et al., 2007); an identical staining pattern has been reported when this antibody was used to detect Müller glia in the adult zebrafish retina (Fausett and Goldman, 2006; Kassen et al., 2007; Kassen et al., 2008; Thummel et al., 2008). In transgenic fish harboring the Müller glia-specific gfap promoter driving EGFP expression, the GS antibody specifically labels EGFP-expressing Müller glia (Kassen et al., 2007).

The PKCβ-1 antibody was made against the C-terminus of human protein kinase C β1 and identifies a 79kDa protein on Western blots from a variety of mammalian cell types (Korchak and Kilpatrick, 2001; Berdiev et al., 2002). In the zebrafish retina, the PKCβ-1 antibody identifies bipolar cells located in the mid to distal region of the INL (Kainz et al., 2003; Nevin et al., 2008; Vitorino et al., 2009). These cells have a characteristic bipolar cell morphology that includes a primary dendrite exiting the distal end of the soma that ends in the outer plexiform layer and an axon emanating from the proximal side of the soma that projects into the inner plexiform layer. Similar to these previous studies (Kainz et al., 2003; Nevin et al., 2008; Vitorino et al., 2009), we also find that the PKCβ-1 antibody identifies retinal cells whose anatomical location and morphology are consistent with bipolar cells.

The zn5 antibody recognizes the activated leukocyte cell adhesion molecule-a (also known as DM-GRASP) that is expressed in zebrafish differentiating motor neurons and retinal ganglion cells (Fashena and Westerfield, 1999; Babb et al., 2005). It does not recognize mature ganglion cells. The characteristic staining of differentiating retinal ganglion cells includes cell bodies located in the GCL and axons extending into the optic nerve layer. This antibody has been used to detect differentiating retinal ganglion cells during zebrafish retina regeneration and these stained cells exhibit the same features as described above (Fausett and Goldman, 2006; Fimbel et al., 2007).

Primary antibodies were detected using secondary donkey anti-rat AMCA (1:250; IgG H + L; Jackson Immunoresearch, West Grove, PA; catalog No. 712-156-150), donkey-anti-rabbit Alexa Fluor 488 (1:1,000; IgG H + L; Invitrogen; catalog No. A-21206) or donkey anti-mouse Alexa Fluor 555 (1:500; Invitrogen; catalog No. A-31570). Omission of primary antibodies resulted in no specific staining.

Quantification of immunostained cells at the injury site

Quantification of GFP, BrdU, HuC/D and GS labeled cells was performed by counting GFP⁺ single positive and GFP⁺/GS⁺, GFP⁺/HuC/D⁺ or GFP⁺/BrdU⁺ double positive cells at 4 injury sites of the retina. Eight retinas were used to quantify cell counts at 2 and 3 days post injury, while two retinas were used to quantify cell counts at 4-7 days post injury. For this analysis the retinal sections we analyzed were 8 microns thick and separated by 40 microns. Only cells with well-defined nuclei were counted. We used the Abercrombie formula to correct for counting errors (Abercrombie, 1946). The mean thickness of sections was 8 microns and the mean nuclear width for GFP⁺, GFP⁺/GS⁺ and GFP⁺/BrdU⁺ cells varied from 5.1 +/-0.34 to 4.9 +/- 0.34 over the time course of the experiment. Ambercrombie's correction for these samples was 0.6. The mean nuclear width for GFP⁺/HuC/D⁺ cells varied from 3.6 +/-0.5 on day 2 post injury to 3.1 +/-0.5 on day 3 post injury and Ambercrombie's correction for these samples was 0.7. Data in Figure 6 are reported as % of GFP⁺ cells co-labeling with GS or BrdU. The raw data and Abercrombie equation corrected data are presented in Supplementary Table 1.

Figure 6. Time course of 4-OHT-dependent recombination in the adult injured retina.

(A) Adult CreER^T2;LCLG transgenic fish were anesthetized and the right retina injured by a needle poke. Twenty hours later fish were immersed in 4-OHT for the indicated times prior to harvesting retina for RT-PCR. The illustration on the left shows the expected sizes of PCR products using the indicated forward and reverse primers (arrows). Shown on the right is an agarose gel of PCR amplified products corresponding to the unrecombined (top band) and recombined (bottom band) message. (B) Quantification of the percentage of GFP⁺ cells that are also GS⁺ or BrdU⁺ at various times following retinal injury. Adult CreER^T2;LCLG transgenic fish retinas were injured and harvested at the indicated days post injury (dpi). Fish harvested at 2-4 days post injury received an IP injection of BrdU 3h before being sacrificed, while fish harvested at 5-7 days post injury received an IP injection of BrdU at 4 days post injury. Retinas were isolated, fixed and sectioned for GFP, glutamine synthetase (GS) and BrdU immunofluorescence detection. (C-Z) Representative confocal images used for quantification reported in panel B. Recombination was restricted to Müller glia (arrows) at early times following retinal injury and at later times recombination is detected in Müller glia (arrows) and other retinal cell types (arrowheads) found in all 3 retinal cell layers. Scale bar in lower right panel = 50 μm (applies to all panels). Abbreviations: BrdU, 5 bromo 2’-deoxyuridine; GFP, green fluorescent protein; GS, glutamine synthetase; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer.

Photomicrograph Production

Microscopic images were imported into Adobe Photoshop CS2 for processing of figures. Contrast and brightness were changed uniformly across the image when necessary. Graphs were prepared in Adobe Illustrator and then imported into Adobe Photoshop for final figure preparation.

Results

Conditional recombination in transgenic fish

Expression vectors (see Materials and Methods and Supplementary Fig. 1) used for lineage tracing in transgenic fish were first tested for conditional gene expression in HEK 293 cells. Co-transfection of HEK 293 cells with 1016 tuba1a:CreER^T2 and β-actin2:LCLG demonstrated 4-OHT-dependent recombination as indicated by GFP expression (Supplementary Fig. 2A-D). 4-OHT did not cause any detectable GFP expression when HEK 293 cells were transfected with only the β-actin2:LCLG recombination reporter (Supplementary Fig. 2E and F).

After establishing that our vectors were functional, we used the Tol2 transposase system (Urasaki et al., 2006) to introduce vector DNA into the zebrafish germ line. 1016 tuba1a:CreER^T2 lines were identified by CFP expression and recombination reporter lines were identified by mCherry expression. We identified both weak and strong expressing 1016 tuba1a:CreER^T2 lines. Weak expressing lines showed CFP expression predominantly in neuromasts that make up the lateral line with weaker expression in the brain (Supplementary Fig. 3A), while strong expressing lines showed expression in neuromasts, spinal cord, brain, and retina (Supplementary Fig. 3B). When we assayed for CreER^T2 RNA expression in the weak expressing lines we detected expression in neuromasts and brain (Supplementary Fig. 3C and D). In general, mCherry expression in the β-actin2:LCLG fish was strong and widespread in developing embryos and adults (Supplementary Fig. 3E-L).

Conditional gene expression during development identifies novel tuba1a expression domains

Double transgenic lines (CreER^T2;LCLG) harboring 1016 tuba1a:CreER^T2 and β-actin2:LCLG transgenes were generated and tested for 4-OHT-dependent recombination. We found that 1016 tuba1a:CreER^T2 lines, that expressed the CFP reporter at high levels, gave background recombination independent of 4-OHT treatment (data not shown) which suggested that CFP expression reflects CreER^T2 levels. Therefore, we focused our studies on double transgenic fish made by crossing a weak expressing 1016 tuba1a:CreER^T2 line (L1) with a strong expressing β-actin2:LCLG line (L19). We found that double transgenic fish tolerated 4-OHT concentrations at and below 2 μM, while higher concentrations were toxic.

To determine if CreER^T2;LCLG fish exhibited 4-OHT-dependent recombination, we immersed fish in 4-OHT or vehicle for 3 days beginning at 1dpf (Fig. 1). Indeed, only fish immersed in 4-OHT exhibited robust panneuronal GFP expression (Fig. 1A and B), which was similar to GFP expression in 1016 tuba1a:GFP transgenic fish (Fig. 1B inset and Fig. 2B).

Figure 1. 4-OHT-dependent recombination in developing CreER^T2;LCLG transgenic fish.

(A, B) CreER^T2;LCLG transgenic fish harboring both the 1016 tuba1a:CreER^T2 and β-act2:LCLG transgenes were incubated with vehicle (-4-OHT) (A) or with 4-hydroxytamoxifen (+4-OHT) (B) at 1dpf and examined under fluorescent microscopy at 4dpf (days post fertilization). Note the panneuronal induction of GFP expression in fish treated with 4-OHT shown in panel (B), while siblings treated with vehicle exhibited undetectable GFP expression shown in panel (A). The inset in panel (B) shows GFP expression in a 1016 tuba1a:GFP transgenic fish for comparison with the expression pattern observed in recombined fish (also see Fig. 2B). Asterisk (*) indicates autofluorescence from the yolk. Scale bar = 100 μm in (A) (applies to B). (C) Agarose gel showing time course of recombination following 4-OHT addition. CreER^T2;LCLG transgenic fish at 24hpf were immersed in vehicle for 48h (C) or 4-OHT for 6, 12, 24 and 48h, before sacrificing fish and harvesting RNA for RT-PCR. The illustration on the left shows the expected sizes of PCR products using the indicated forward and reverse PCR primers (arrows). The forward primer binds to untranslated sequence in exon 1 from the β-actin2 gene and the reverse primer binds to the GFP sequence. Shown on the right is an agarose gel of PCR amplified products corresponding to the unrecombined (top band) and recombined (bottom band) message. Note that within 6h of 4-OHT application significant amounts of the recombined mRNA can be detected and that by 48h of 4-OHT application recombination has leveled off. Stnd are 1kb DNA molecular weight standards.

Figure 2. Temporal control of recombination during development reveals transient 1016 tuba1a promoter activity in neural and non-neural cell populations.

(A) CreER^T2;LCLG transgenic fish were immersed in 4-OHT for 6h starting at 13, 24, 48, and 72hpf and assayed for GFP expression at 7dpf. Note the robust and panneuronal expression in fish immersed in 4-OHT at 24hpf or earlier. (B) GFP expression in 1016 tuba1a:GFP transgenic fish at 7dpf for comparison with recombined expression pattern. (C, D) 4-OHT-dependent recombination in developing muscle (arrow) of CreER^T2;LCLG transgenic fish immersed in 4-OHT at 13 and 24hpf. (E) Lack of GFP expression in muscle of 1016 tuba1a:GFP fish at 7dpf. (F, G) 4-OHT-dependent recombination in heart cells is only observed when fish are immersed in 4-OHT at 13hpf. (H) Lack of GFP expression in the heart of 1016 tuba1a:GFP fish. (I, J) 4-OHT-dependent recombination in the gut of fish immersed in 4-OHT at 13hpf and 24hpf. (K) Absence of GFP expression in the gut of 1016 tuba1a:GFP fish. Asterisk (*) indicates autofluorescence from the yolk. Scale bar = 100 μm in A (applies to B), C (applies to D-E) and F (applies to G-K).

To determine if shorter times of exposure to 4-OHT would also induce recombination; we used RT-PCR to assay recombination at 6, 12, 24, and 48h post 4-OHT immersion (Fig. 1C). Sibling control fish incubated in vehicle were also harvested at 48h post vehicle immersion and assayed for recombination (Fig. 1C, lane C). Although recombination was not detected in vehicle-treated fish, immersion in 4-OHT for as little as 6h was sufficient to induce recombination (Fig. 1C). DNA sequence analysis of the 1.4kb and 0.67kb bands confirmed they represented the original and recombined transcripts, respectively. DNA sequencing showed the very faint band at about 1kb represents the recombined product where the reverse primer bound to an imperfect complementary sequence about 300 nucleotides downstream of its intended target.

We next used CreER^T2;LCLG transgenic fish to investigate when and where the 1016 tuba1a promoter is activated during development. For these experiments transgenic fish received a 6h pulse of 4-OHT at 13, 24, 48, 72 and 96hpf and GFP expression was recorded at 7dpf (Fig. 2A). 1016 tuba1a promoter activity, reflected by GFP expression, was most robust in fish pulsed with 4-OHT at 13 and 24hpf, and is consistent with this promoter being activated panneuronally in neural progenitors. This pattern of expression was similar to that of 1016 tuba1a:GFP fish (Fig. 2B); however, 1016 tuba1a:GFP fish exhibited a more restricted pattern of neural GFP expression in the brain and spinal cord than did CreER^T2;LCLG fish (Fig. 2A and B, Supplementary Fig. 4). This difference in expression can largely be explained by the persistent versus transient nature of the β-actin2 and the 1016 tuba1a promoters, respectively.

We were surprised to find transgene expression in heart, intestine and skeletal muscle in CreER^T2;LCLG fish pulsed with 4-OHT at 13-24hpf that was not apparent in 1016 tuba1a:GFP fish (Fig. 2C-K). This suggested a transient activation of the 1016 tuba1a promoter in cells giving rise or contributing to these non-neural tissues that is undetectable in 1016 tuba1a:GFP fish. To determine if the non-neuronal pattern of 1016 tuba1a promoter activity reflected that of the endogenous tuba1a gene we assayed expression of the endogenous gene by in situ hybridization at 24hpf. This analysis showed that the vast majority of the tuba1a mRNA is expressed in the developing nervous system (Fig. 3); however overdevelopment of the in situ signal revealed tuba1a mRNA in caudal somites that generate muscle (Fig. 3, arrows) (Devoto et al., 1996) and endoderm that gives rise to the gut (Fig. 3, arrowheads) (Ober et al., 2003). No in situ hybridization signal was detected, even after long exposures, with a control sense-strand probe (Supplementary Fig. 5). Thus this conditional recombination system provides a sensitive and robust readout of weak and transient tuba1a promoter activity in previously unappreciated expression domains. Although we have not investigated the significance of this expression, the temporal expression pattern suggests tuba1a contributes to progenitor proliferation and differentiation in these non-neural tissues. These results illustrate the utility of using a conditional recombination system that allows one to probe for weak and transient expression throughout development that would be missed using more traditional approaches and provides a means for labeling these cells or manipulating them by conditional expression of transgenes.

Figure 3. In situ hybridization reveals endogenous tuba1a gene expression in gut endoderm and somites.

Wildtype fish at 24hpf were fixed and hybridized with a digoxigenin antisense tuba1a RNA probe. Note high level expression of tuba1a RNA in the brain (B) and spinal cord (SC). Arrows point to reduced expression in the caudal somites and arrowheads point to expression in the gut endoderm. Scale bar = 100 μm.

Conditional gene expression in adults and following retina injury allows lineage tracing of Müller glia-derived retinal progenitors

After establishing that recombination and reporter gene expression in CreER^T2;LCLG transgenic embryos and fry were inducible by 4-OHT, we investigated if recombination could be regulated by 4-OHT in adult fish. The simplest and most reproducible method for inducing widespread recombination with minimal lethality was immersion in 4-OHT; However, for recombination in the eye, immersion in 4-OHT or eye injection of 4-OHT worked equally well.

Control CreER^T2;LCLG fish exhibited very little to no recombination as assayed by GFP expression in live animals (Fig. 4A and B). An example of the type of background we observe in about 25% of the CreER^T2;LCLG transgenic fish in the absence of 4-OHT is shown in Fig. 4A. In contrast, CreER^T2;LCLG fish immersed in 4-OHT for 3 days and examined 2 months later exhibited GFP expression in olfactory pits, brain and the anterior and posterior lateral line (Fig. 4C and D). The pattern of 4-OHT-induced transgene expression in CreER^T2;LCLG fish was similar to that of the 1016 tuba1a:GFP fish (Fig. 4E and F), although GFP expression in the anterior lateral line was reduced in 1016 tuba1a:GFP fish. Confocal microscopy of tissue sections from 4-OHT-treated CreER^T2;LCLG fish confirmed that the 1016 tuba1a promoter was expressed in pear-shaped sensory hair cells of the lateral line neuromast and putative sensory cells located in the apical portion of the olfactory epithelium (Fig. 4G and H). Consistent with the observation that cells in these anatomical locations do not normally proliferate (Williams and Holder, 2000; Byrd and Brunjes, 2001), we were unable to co-label them with a 3 hr pulse of BrdU (data shown for olfactory epithelium in Fig. 4H).

Figure 4. 4-OHT-dependent recombination in adult CreER^T2;LCLG transgenic fish.

(A-D) Adult CreER^T2;LCLG transgenic fish approximately 5mo old were immersed in vehicle (-4-OHT) or 4-OHT (+4-OHT) for 3 days and GFP expression visualized using a stereo fluorescence microscope. (E, F) GFP expression in 1016 tuba1a:GFP transgenic fish is shown for comparison. Panels A, C and E are dorsal views. Panels B, D and F are lateral views. (A) Arrow points to rare GFP⁺ cell in vehicle-treated fish. (C-F) Arrows point to GFP⁺ cells in olfactory pits, neuromasts and midbrain. Scale bar = 1mm in panel F and applies to panels A-E. (B) Confocal microscopy of cryosections through the lateral line neuromast (G) and olfactory pit (H) shows GFP expression in putative sensory cells (arrows). The fish in panel H was also exposed to BrdU for 2 days following 4-OHT treatment. BrdU⁺ nuclei are red/purple color and marked by an arrowhead. Confocal microscopy suggests that GFP⁺ sensory cells, which are concentrated in the apical olfactory epithelium, do not co-label with BrdU. (*) Identifies non-specific fluorescence in the lumen of the olfactory epithelium. Scale bars are 10 μm in (G) and 50 μm in (H).

1016 tuba1a:GFP fish have been instrumental in identifying Müller glia-derived retinal progenitors that facilitate regeneration of an injured retina (Fausett and Goldman, 2006). However, because 1016 tuba1a:GFP transgene expression does not persist in these putative progenitors after they exit the cell cycle and differentiate, it remained unclear whether newly regenerated cells were derived from the 1016 tuba1a:GFP expressing Müller glia. Here we investigated if the CreER^T2/LoxP recombination system could be used to trace the lineage of Müller glia-derived progenitors that re-entered the cell cycle.

To visualize injury and 4-OHT-dependent recombination in the retina, we injured retinas of double transgenic fish with a needle poke and immersed fish in 4-OHT for 3 days. Controls received vehicle in place of 4-OHT. Fish then received an IP injection of BrdU on day 4 post injury. Fish were sacrificed 2 weeks post-injury and retinas were processed for GFP and BrdU immunohistochemistry. Except for variable GFP expression in a few cells around the optic nerve head (observed in about 50% of the transgenic fish), there was no GFP expression in the uninjured retina regardless of 4-OHT treatment (Fig. 5A-F). A few BrdU-positive rod progenitors were observed in the outer nuclear layer especially in the retinal periphery (Fig. 5B and E). In contrast, following retinal injury, only fish that received 4-OHT exhibited GFP expression at the injury site (which was visualized by a large increase in BrdU-labeled cells) (Fig. 5G-L).

Figure 5. 4-OHT-dependent recombination in the injured retina is restricted to the injury site.

Adult CreER^T2;LCLG transgenic fish were anesthetized and retinas left uninjured (A-F) or injured by a needle poke (G-L). Twenty hours later fish were immersed in vehicle (-4-OHT, panels A-C and G-I) or 4-OHT (+4-OHT, panels D-F and J-L) for 3days, and then received an intraperitoneal injection of BrdU on day 4 post injury and returned to their tank for 2 weeks before harvesting retinas. Retinas were sectioned and stained with DAPI (A, D, G, J), anti-BrdU (B, E, H, K) and anti-GFP (C, F, I, L) antibodies, to identify proliferating and recombined cells, respectively. (A-C) Uninjured retina shows a single BrdU-positive rod progenitor, but no GFP expression. (D-F) Uninjured retina treated with 4-OHT shows a few BrdU expressing rod progenitors, but no GFP expression. (G-I) Injured retina shows a large induction of BrdU expressing cells at the site of injury (*), but no GFP expression. (J-L) Injured retina treated with 4-OHT shows a large induction of BrdU expressing cells at the site of injury (*) and an increase in GFP expressing cells. Scale bar = 100 μm in L (applies to A-K).

We next used RT-PCR to assay recombination in the injured retina of CreER^T2;LCLG fish (Fig. 6A). For this analysis, retinas were injured by a needle poke and 1 day later fish were immersed in 4-OHT for up to 3 days. Retinas were harvested beginning at the time when fish were first immersed in 4-OHT (0 days post 4-OHT immersion) and then at daily intervals for up to 14 days (3 retinas per time point). Retinal RNA was isolated and used in RT-PCR reactions with a β-actin2 exon 1 forward primer and a GFP reverse primer (arrows in Fig. 6A). This analysis showed that expression of the recombined message began around 2 days post 4-OHT immersion. In general recombination in the uninjured retina was undetectable, even after exposure to 4-OHT, or in the injured retina of fish that received vehicle devoid of 4-OHT (Fig. 6A). However periodically we observed a very faint band corresponding to recombined message in the uninjured retina (for example see Fig. 6A, Uninjured 4 days post 4-OHT immersion) which we suspect represents the variable and small amount of recombination observed near the optic nerve head. When comparing 7 with 14 days post 4-OHT immersion it appears there is a slight reduction in the recombined 670bp band which likely reflects a less severely injured retina used for that time point. DNA sequence analysis of the unrecombined 1.4kb band and the recombined 670bp band indicated that RNA processing and recombination occurred without any errors (data not shown). The 1.4kb band is present in all samples because the RNA used for RT-PCR was from whole retina where most cells express unrecombined reporter (1.4kb band) but do not express CreER^T2 and hence exhibit no recombination. These experiments were repeated 3 times with similar results.

We had previously shown the 1016 tuba1a promoter is activated in Müller glia at early times following retinal injury (Fausett and Goldman, 2006). To determine if recombination is also restricted to these cells, we co-stained injured retinas with anti-GFP and anti-GS antibodies (Fig. 6C-Z). Consistent with the RT-PCR analysis, immunohistochemical detection of GFP was first observed on day 2 post retinal injury and the vast majority of these cells co-stained with the Müller glia marker, GS (Fig. 6C-F and Supplementary Table 1). Examination of 8 injured retinas indicated that of approximately 3,725 GFP⁺ cells identified at 2 and 3 days post injury, 3501 (94%) of these were also GS⁺ (Fig. 6B and Supplementary Table 1). Using cell-type-specific antibodies, we found that the remaining GFP⁺/GS^- cells expressed the HuC/D antigen, exhibited extensive dendritic branching, and were localized to the proximal portion of the INL (Supplementary Fig. 6) which suggests they are amacrine cells (Godinho et al., 2005). In addition, these cells did not reenter the cell cycle, since they could not be labeled with BrdU. Therefore, to unambiguously show 1016 tuba1a-expressing Müller glia-derived progenitors give rise to amacrine cells, additional methods must be employed; such as conditional gene expression in combination with BrdU labeling (see below and Fig. 7).

Figure 7. Tuba1a-expressing progenitors regenerate retinal neurons and glia.

Adult double transgenic fish retinas were injured by a needle poke and then fish were exposed to 4-OHT either by immersion or direct injection into the eye. On day 4 post injury fish were injected intraperitoneally with BrdU and allowed to survive for 1-2.5 months. Retinas were then harvested, sectioned and stained with anti-GFP, anti-cell-type-specific (zpr1 for photoreceptors in the ONL, GS for Müller glia, PKC for bipolar cells, and HuC/D for amacrine cells in the INL, and zn5 for ganglion cells in the GCL) and anti-BrdU antibodies. Sections were analyzed using confocal microscopy. Arrows point to triple-labeled cells. Scale bar in panel D (applies to A-C), panel H (applies to E-G), panel P (applies to panels M-O) and panel T (applies to panels Q-S) is 20 μm; Scale bar in panel L (applies to panels I-K) is 15 μm. Abbreviations: BrdU, bromodeoxyuridine; GFP, green fluorescent protein; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer.

At later times post injury we noticed a gradual decrease in the percentage of GFP⁺ cells that were also GS⁺ (Fig. 6B and Supplementary Table 1) and an increase in the number of GFP⁺ cells found in all retinal layers (Fig. 6W-Z arrowheads). This is compatible with the idea that GFP⁺ Müller glia give rise to other retinal cell types that contribute to repair of the damaged retina.

Müller glia-derived progenitors reenter the cell cycle around day 2 post retinal injury, exhibit peak proliferation at day 4 post injury and return to baseline levels by day 7 post injury (Fausett and Goldman, 2006). To investigate if GFP⁺ cells were proliferating Müller glia-derived progenitors, CreER^T2;LCLG fish retinas were injured and exposed to 4-OHT and then, 3h prior to harvesting, fish received an IP injection of BrdU. At 3 days post injury, we found that approximately 5% of the GFP⁺ cells were also BrdU⁺ which increased to approximately 15% on day 4 post injury (Fig. 6B and G-N). Some of the fish that received BrdU on day 4 post injury were allowed to develop further (up to 7 days post injury) and the percentage of GFP⁺ cells that were also BrdU⁺ were quantified. This percentage remained relatively constant (Fig. 6B) and suggests that the 1016 tuba1a promoter may be most active in Muller glia-derived progenitors that are entering their last cell division and preparing to differentiate. Because we are selecting very low expressing 1016 tuba1a:CreER^T2 lines to prevent ligand-independent recombination, we may only be able to observe recombination in these late stage progenitors. Consistent with this idea is our observation that only 5-10% of the BrdU⁺ cells in an injured retina express GFP at 4 days post injury. In contrast, most BrdU⁺ cells express GFP in injured retinas of 1016 tuba1a:GFP fish (Fausett and Goldman, 2006), which likely reflects selection for high promoter activity that results in GFP expression at early and late progenitor stages.

To determine if tuba1a expressing progenitors contribute to regeneration of mature retinal neurons and glia following retinal injury we took advantage of our CreER^T2;LCLG fish to permanently label the tuba1a-expressing Müller glia with GFP. We also labeled dividing cells by injecting BrdU IP into fish on day 4 post injury when cell proliferation is maximal (Fausett and Goldman, 2006). At 10 days to 2.5 months post retinal injury, fish were sacrificed and retinas were sectioned and stained for GFP, cell-type specific markers and BrdU, and analyzed by confocal microscopy. GFP expression serves as a lineage tracer for descendents of tuba1a-expressing progenitors, cell-type-specific antibodies allow for identification of individual retinal cell types, and BrdU-labeling confirms cells were derived from cycling Müller glia-derived progenitors. We found tuba1a-expressing progenitors gave rise to photoreceptors in the ONL that expressed the Zpr1marker of red/green cones (Fig. 7A-D), and also gave rise to Müller glia (Fig. 7I-L), bipolar cells (Fig. 7E-H) and amacrine cells (Fig. 7M-P) that reside in the INL and were identified by the co-expression of cell-specific markers GS, PKC and HuC/D, respectively. At 10 days post injury when newly differentiating cells can still be identified, we found tuba1a-expressing progenitors gave rise to retinal ganglion cells expressing the developmental marker Zn5 (Fig. 7Q-T).

Discussion

The tuba1a promoter has traditionally been viewed as a neural specific promoter that is induced during nervous system development and differentiation (Heiber et al., 1998; Goldman et al., 2001). However, this promoter is also active in adult neural progenitors lining brain ventricles, the central canal of the spinal cord and the periphery of the retina (Goldman et al., 2001). In the adult retina the tuba1a promoter is activated, following injury, in Müller glia-derived retinal progenitors (Fausett and Goldman, 2006). These progenitors are thought to be responsible for regenerating a damaged retina in zebrafish. In mammals, Müller glia are much more restricted in their response to retinal injury; regenerating only a very limited number of retinal cell types with very low frequency even when stimulated by growth or differentiation factors (Ooto et al., 2004; Karl et al, 2009).

To better visualize these tuba1a-expressing cells during CNS development and regeneration, we previously created transgenic fish harboring the 1016 tuba1a:GFP transgene (Fausett and Goldman, 2006). This transgene contains a fragment of the tuba1a promoter that appears to restrict transgene expression to the nervous system during development and to Müller glia-derived progenitors in the adult injured retina. Although 1016 tuba1a:GFP transgenic fish have facilitated our studies of CNS development and regeneration by marking Müller glia-derived retinal progenitors in the injured retina, they do not allow us to follow the fate of these progenitors since transgene expression is extinguished as they differentiate.

To facilitate visualization of tuba1a progenitors and their fate, we generated double transgenic fish harboring 1016 tuba1a:CreER^T2 and β-actin2:LCLG transgenes (Fig. 1). These fish allow us to take advantage of the power of conditional gene expression and recombination to specifically and permanently label tuba1a-expressing cells in the developing and adult CNS. Recombination during early stages of development showed that tuba1a-expressing progenitors contribute to not only the nervous system, but also to heart, skeletal muscle and intestine (Fig. 2). This was a surprising result since previous analysis of 1016 tuba1a:GFP transgenic fish suggested this promoter was restricted to neural tissue (Fausett and Goldman, 2006). We reason that this non-neural expression escaped detection in previous studies because the tuba1a promoter is expressed transiently and at a low level in cells that contribute to development of heart, skeletal muscle and intestine. In contrast, even very low levels of CreER^T2 expression in 1016 CreER^T2;LCLG fish that were selected for weak expression would stimulate recombination and maintain strong GFP expression driven by the β-actin promoter and thus facilitate visualization of these non-neural cells.

By immersing embryos in 4-OHT at different times during development we were able to identify a developmental window (13-24hpf) when this non-neural expression was occurring (Fig. 2). In situ hybridization assays confirmed our lineage tracing studies and demonstrated very low, but detectable expression of the endogenous tuba1a gene in these non-neural tissues at 24hpf (Fig. 3). It is interesting that an alpha tubulin isoform, like tuba1a, that is predominantly expressed in the developing nervous system is also expressed in a limited number of non-neural tissues during early development. Although the reason for expressing different tubulin isoforms in different tissues is not known, our data suggest that the unique function it imparts to the developing nervous system may also be shared with cells contributing to heart, skeletal muscle and intestine. These studies also indicate that the conditional recombination system used here is a very sensitive readout of gene expression that allows one to not only trace cell lineages, but also define developmental windows of gene expression that can be visualized in live animals. Finally, our studies suggest that the 1016 tuba1a promoter can be used to label or manipulate these non-neuronal progenitors and their progeny using conditional gene expression.

Adult 1016 tuba1a:GFP fish express GFP in neuromasts of the lateral line, olfactory pits and restricted brain regions. As expected, a similar pattern of expression is observed following immersion of CreER^T2;LCLG fish in 4-OHT (Fig. 4). BrdU labeling and immunohistochemistry of sectioned material showed that in the lateral line and olfactory epithelium, recombination is restricted to non dividing sensory cells (Fig. 4G and H).

We previously showed that in 1016 tuba1a:GFP fish, retinal injury transiently induces GFP expression in a cycling population of Müller glia-derived retinal progenitors (Fausett and Goldman, 2006). However, because GFP expression is transient in these fish we were not able to directly demonstrate that Müller glia expressing the 1016 tuba1a promoter were the same cells that regenerated the retina following injury. This is also true for other transgenic models using promoters that are active either in Müller glia or retinal progenitors, but shut off in differentiated retinal neurons (Fausett and Goldman, 2006; Bernardos et al., 2007; Fimbel et al., 2007; Thummel et al., 2008). Thus, there were no lineage tracing studies demonstrating Müller glia-derived retinal progenitors gave rise to retinal neurons or glia that integrate into the retinal architecture and were stably maintained. To address this issue we created 1016 CreER^T2;LCLG fish to permanently label Müller glia-derived retinal progenitors in the injured retina.

Analysis of recombination at 2-3 days post retinal injury indicated that 94% of the recombined GFP⁺ cells could be identified as Müller glia, while the remainders were identified as amacrine cells (Supplementary Table 1). We never detected GFP expression in amacrine cells following retinal injury in our 1016 tuba1a:GFP transgenic fish, suggesting that 1016 tuba1a promoter activity may be too low and/or transient to drive enough GFP expression for detection. However, this low and transient promoter activity may be sufficient to induce recombination in 1016 CreER^T2;LCLG fish which, following recombination, would express GFP constitutively from the strong β-actin promoter and therefore allow easy detection. Thus, similar to what we reported in our lineage tracing experiments during development, the recombination system allows for a more sensitive detection of low and transient 1016 tuba1a promoter activity that may be missed using more traditional transgenic approaches. Although this low level of recombination in amacrine cells precluded us from solely using recombination to trace the amacrine lineage, we were able to combine BrdU-labeling with lineage tracing to clearly show that tuba1a-expressing Müller glia-derived progenitors regenerate amacrine cells along with most other retinal cell types (Fig. 7).

Quantification of recombination and cell proliferation at 4 days post injury suggested that only 5-10% of the BrdU⁺ cells underwent recombination in CreER^T2;LCLG fish, while the majority of BrdU⁺ cells express GFP in 1016 tuba1a:GFP fish (Fausett and Goldman, 2006). This latter observation suggests the 1016 tuba1a promoter is normally active in the proliferating cell population, but the former data suggest its highest activity is restricted to a subpopulation of proliferating cells. Indeed, a BrdU pulse/chase experiment suggested that these cells are late-stage retinal progenitors that are undergoing their final divisions (Fig. 6B). Thus by selecting a very weak expressing 1016 tuba1a:CreER^T2 line to prevent basal ligand-independent recombination, we may have restricted recombination to these late-stage retinal progenitors. In addition, recombination itself may be reduced by low level CreER^T2 expression as reported in mice (Feil et al., 1996).

Regarding the transgenic lines we generated; our high and ubiquitous expressing β-actin2:LCLG line should provide investigators with a universal reporter for Cre activity, while our analysis of different 1016 tuba1a:CreER^T2 lines suggests that only low CreER^T2-expressing lines will be useful for conditional expression experiments. This latter observation, which is similar to what we found in transfected tissue culture cells, suggests that CreER^T2 may escape sequestration when expressed at high levels. Indeed CreER sequestration may be enhanced and ligand-independent recombination reduced by appending an additional ER^T2 domain (Boniface et al., 2009; Matsuda and Cepko, 2007).

One goal of generating a conditional gene expression system using CreER^T2/LoxP technology, in addition to lineage tracing, is to be able to induce genes in a specific cell type at any time during an animal's life cycle and examining the consequence of that gene on cell function. The conditional recombination system described here is ideally suited for this purpose. By choosing appropriate promoters to drive CreER^T2 expression and tagging the conditionally expressed protein (for example generating fusions with fluorescent proteins) one can direct expression to any cell type at any time of development and follow the cell's fate. Because of the mosaic nature of the conditional recombination system, one should be able to follow these cells and compare them to their normal genetically unmodified neighbors for changes in behavior and function.

In conclusion, we have shown that the CreER^T2/LoxP system applied to zebrafish enables conditional gene expression and lineage tracing in developing and adult zebrafish. We used this system to identify the fate of tuba1a-expressing cells during development and during retinal regeneration. Our data suggest that during development, tuba1a-expressing progenitors not only contribute to neural tissue, but also heart, muscle and intestine. In the adult injured retina, our data suggest that tuba1a-expressing Müller glia-derived progenitors are responsible for regenerating most retinal neurons and glia and that these cells are stably integrated into the retinal architecture. Conditional gene expression and lineage tracing in zebrafish will open up new avenues for studying development and regeneration.