The definitive endoderm forms during gastrulation and has both structural and instructive roles during embryogenesis. Many visceral organs, including the liver, pancreas, pharyngeal glands, and the epithelia of the respiratory tract are derived from the primitive gut tube (Tam et al.,2003). In addition to providing requisite inductive cues after gastrulation for correct head formation and initiation of heart morphogenesis (Tam et al.,2003), the endoderm is an ongoing source of critical patterning signals to these and other tissues (Tam et al.,2003; Fukuda and Kikuchi,2005). Likewise, the notochord and floorplate of the neural tube provide required patterning signals to adjacent tissues, including endoderm, and to one another (Cleaver and Krieg,2001; Wilson and Maden,2005).
Temporospatially regulated inactivation of genes provides a powerful approach to elucidate the genetic networks that function within these tissues and to uncover their signaling properties over developmental time. We previously reported a tet-on system for inducible expression of Cre recombinase in endoderm from the Foxa2 locus (Foxa2Ires rtTA and TRE Cre; Frank et al.,2007). This system provides reliable Cre activity in endoderm, notochord, and floorplate as well as in gut derivatives and circumvents the embryonic lethality that can result from inactivation of genes required for early embryogenesis (Frank et al.,2007). However, this system is genetically complex as it requires the accumulation of five transgenic alleles to generate a conditional mutant embryo: two conditional alleles of the gene of interest, the transgene containing the tet-response element driving Cre recombinase expression, and homozygosity for the Foxa2Ires rtTA allele. Here, we present an improved Cre-expression system at the Foxa2 locus that is rapidly responsive to tamoxifen induction and genetically less cumbersome. We also provide new insight into the optimal dosage, timing, and route of administration of tamoxifen to maximize Cre activity while minimizing embryonic toxicity.
RESULTS AND DISCUSSION
Generating the Foxa2mcm Allele
Nuclear translocation and recombinase activity of a fusion protein containing Cre recombinase flanked by mutated hormone-binding domains of the murine estrogen receptor (mER;Cre;mER; Verrou et al.,1999) is dependent on binding by tamoxifen in the cytoplasm (Zhang et al.,1996). We took advantage of the temporally inducible nature of this system and combined it with the tissue-specific and temporal properties of the Foxa2 locus to provide dual axes for regulating Cre-recombinase activity and to thus control the activity of loxP-flanked (floxed) alleles within Foxa2 expression domains (Ang et al.,1993,1996; Monaghan et al.,1993; Kinder et al.,2001).
With the goal of obtaining mER;Cre;mER production in Foxa2-expressing tissues without disrupting Foxa2 function, we generated a cassette containing an IRES (Jackson et al.,1990; Jang et al.,1990) followed by sequence encoding the mER;Cre;mER (Verrou et al.,1999), and an frt-flanked neomycin phosphotransferase gene (neor) for selection of targeted ES clones (Fig. 1). We flanked the neor with frt sites for subsequent removal with Flp-recombinase in founder animals with the goal of avoiding hypomorphic effects of neor on function of Foxa2 or other loci (Olson et al.,1996; and demonstrated herein).
To determine whether this novel Foxa2mcm allele has intact Foxa2 function after removal of neor with Flp recombinase, we examined phenotypes and Foxa2 mRNA expression in animals hetero- and homozygous for Foxa2mcm. Foxa2 mRNA expression in Foxa2mcm/+ and Foxa2mcm/mcm embryos was detected in the rostral floorplate, foregut, and node at embryonic day (E) 8.5 and the expression pattern and level was indistinguishable in heterozygote and homozygote Foxa2mcm embryos and wild-type controls (Fig. 2A–C). These embryos develop normally without phenotypes characteristic of Foxa2 null, heterozygous null, or conditional mutant embryos (early embryonic lethality, neural and endodermal differentiation defects; Ang and Rossant,1994; Weinstein et al.,1994). The adults are indistinguishable from wild-type animals and reproduce normally (data not shown). These results indicate that the presence of the Iresmcm cassette does not adversely affect Foxa2 function. In contrast, embryos homozygous for the neor-containing form of the allele (Foxa2mcm-neo/mcm-neo) have developmental delay, abnormal heart and head development, and embryonic lethality, indicating that the Foxa2mcm-neo allele is hypomorphic (Fig. 2D–F).
Inducible Cre Activity From the Foxa2mcm Allele
To analyze tamoxifen-inducible Cre recombinase activity from this allele, Foxa2mcm mice were crossed with Rosa26LacZ reporter mice (Soriano,1999) and a single dose of tamoxifen was administered to pregnant females. The embryos were assessed by X-gal staining at different somite stages (ss) after exposure to 0.05 mg/g maternal body weight (bw) of tamoxifen at E5.75 by intraperitoneal (IP) injection. At 0 ss, X-gal staining was detected in the definitive endoderm (Fig 3A,A′). LacZ-positive cells were also observed in extraembryonic tissues likely derived from definitive visceral endoderm (DVE) and anterior visceral endoderm (AVE; Fig. 3A,A′). Foxa2 is expressed in the DVE at E5.5 and in the AVE at E5.75 to E6.0 (Ang and Rossant,1994; Weinstein et al.,1994); thus, Cre protein present in these domains can be translocated into the nucleus when tamoxifen is administrated. At the 0 ss stage, we also reproducibly detected a few LacZ-positive cells in anterior mesoderm, including cardiac precursors in the cardiac crescent (Fig. 3A–A″). Because Foxa2 mRNA expression has been reported in the anterior mesoderm of the embryo at late streak and early bud stage (Kinder et al.,2001), it is likely that the LacZ-expressing cells we detect in this region represent this lineage. At E8.5 (10 ss), foregut endoderm, notochord, and floorplate were LacZ-positive (Fig. 3B,B′). Cre activity was also reproducibly detected in a small subset of head mesenchymal and myocardial/endocardial cells. Notably, we did not detect cardiac staining in embryos dosed after E7.5 (Fig. 4C,D), consistent with these labeled cells deriving from Foxa2-expressing cells at the late streak and early bud stage. At E9.5, lacZ-positive cells continue to reflect the Foxa2 lineage and include pharyngeal and gut endoderm, liver bud, notochord, and floorplate (Fig. 3C, C′). Overall, the activity of Cre-mediated recombination indicates that the Foxa2mcm allele recapitulates endogenous Foxa2 expression.
For rigorous temporal regulation of Cre activity within the Foxa2 domain, it is important that there is no Cre activity in the absence of tamoxifen. In embryos bearing the Foxa2mcm and Rosa26LacZ alleles, we observed no X-gal staining in the absence of tamoxifen administration and no recombined alleles could be detected in polymerase chain reaction (PCR) assay of tail or yolk sac DNA samples (data not shown).
Influence of Dose, Timing, and Method of Tamoxifen Administration Affect Cre Activity
To examine the effects of tamoxifen dosage and timing of administration on induction of Cre activity, we varied these parameters in pregnant females and examined the resulting Cre activity. After administration of tamoxifen IP at 0.025 mg/g bw at E6.5, patchy staining was detected in the Foxa2 expression domain at E9.5, whereas administration of 0.05 mg/g bw at E6.5 resulted in more uniform staining (Fig. 4A,A′,B,B′). Relative to its expression at earlier embryonic stages, Foxa2 expression is decreased in the definitive endoderm at E8.0 (Burke and Oliver,2002; Frank et al.,2007), and this finding explains the more variable Cre activity detected in definitive endoderm-derived tissues after tamoxifen dosing at E7.5 (Fig. 4C,C′). By delaying tamoxifen administration to E8.5, we could isolate activity to the Foxa2mcm floorplate expression domain (Fig. 4D,D′). These results indicate that the level and domain of Foxa2mcm activity can be regulated within the Foxa2 domain by size of tamoxifen dose and timing of administration.
To maximize the rapidity of onset and extent of Cre activity within the desired expression domain, while minimizing drug toxicity, we investigated the route and dosing protocols for tamoxifen administration. In rats, tamoxifen administered by oral gavage achieves a peak concentration in blood at 6 hr and quickly decays to basal levels within 12 hr (Shin et al.,2006). Thus, we postulated that oral gavage would rapidly induce Cre activity and have less toxicity than we observe with IP injection in mice. We compared toxicity and Cre activity obtained with IP vs. oral gavage with Foxa2mcm. The maximum dose of tamoxifen that could be given IP at E6.5 was 0.05 mg/g bw because at 0.0625 mg/g bw, we frequently noted developmental delay and abnormal head development. In contrast, oral gavage at doses of 0.1 or 0.12 mg/g bw at E6.5 and E7.5 were well tolerated, and the embryos were indistinguishable from untreated controls (data not shown). Although there were no developmental defects in the embryos, we observed intrauterine hemorrhage (four of five) at E18.5 with 0.12 mg/g bw tamoxifen administration at E6.5. This hemorrhage was not observed at E17.5, and the embryos appear healthy and normal. Thus, 0.12 mg/g bw of tamoxifen can be used without perturbing embryogenesis. In terms of comparing Cre activity, females were treated with either 0.05 mg/g maternal bw IP or 0.12 mg/g bw oral at E6.75 and embryos were harvested at E8.5 and E9.5 to examine X-gal staining. We found that recombination by Cre obtained after oral gavage was more uniform and robust at all stages assayed. Furthermore, the intensity of X-gal staining in recombined tissues obtained after gavage was greater, revealing that more β-galactosidase and its reaction product had accumulated in the cells after administration (Fig. 5). This finding indicates that the onset of Cre activity obtained with oral gavage is more rapid than with IP.
To specifically test the onset of Cre activity post-tamoxifen gavage, we examined induction of Cre activity produced by the globally expressed, tamoxifen inducible RosaCreEsR allele (Badea et al.,2003), Rosa26LacZ reporter females were bred to RosaCreEsR allele bearing males and then gavaged with tamoxifen at E6.5. Embryos were stained for X-gal activity at approximately 6, 12, 15, 38 hr after gavage. We detected recombination in some cells by 6 hr post-administration and evidence of global activation by 12 hr (Fig. 6). More rapid onset of Cre activity was obtained with the oral dose of 0.12 mg/g bw than 0.08 mg/g bw.
In summary, this study reports the generation and characterization of a tamoxifen-inducible Cre driver active in Foxa2 expression domains that can be used for lineage analyses and temporospatial regulation of gene function in endoderm-, notochord-, and floorplate-derived tissues. The Foxa2mcm allele will be a valuable reagent to study the functions of endoderm, notochord, and floorplate derived signals and transcription factor networks. The advantages of this system compared with the Foxa2ITA-TetOCre system (Frank et al.,2007) we previously reported are that the Foxa2mcm system is less genetically cumbersome and more easily titratable in terms of controlling timing and location of Cre activity. One allele of Foxa2mcm is sufficient to induce Cre activity throughout the Foxa2 expression domain at a given time. We have also demonstrated that administration of tamoxifen by oral gavage provides superior Cre activity, more rapid onset, and less toxicity than obtained with IP dosing.
Design and Construction of the Foxa2mcm Allele
We inserted an IRES (Jackson et al.,1990; Jang et al.,1990) in frame upstream of a merCremer;neor cassette (generously provided by Michael Reth; Verrou et al.,1999). This cassette was inserted into a Bst11071 site located 203 bp 3′ to the Foxa2 stop codon and 5′ to the endogenous polyadenylation sequence (Frank et al.,2007; Fig. 1). Homology included 8,678 bp 5′ to the Bst11071 and 2,690 bp 3′ of the insertion. The modified genomic fragment was shuttled into a plasmid backbone flanked by thymidine kinase genes. The linearized targeting vector was electroporated into ES cells and subjected to selection. Positive clones were identified by digesting genomic DNA with HindIII, Southern blotting, and hybridizing with a 3′ flanking probe, and targeted clones were confirmed correct by additional analyses with internal genomic probes and probes within the cassette. Genotyping for the targeted allele was performed using PCR on tails or yolk sacs using following primers: 5′ CTCAAGGGAGCAGTCTCACC (forward), 5′GA CTTTTCTGCAACAACAGCA (reverse), and 5′ATACTATCTAGAGAATAGGAACTTCG (frt, forward) were multiplexed to identify wild-type and targeted alleles.
Tamoxifen Treatment of Mice
Tamoxifen (Sigma, T5648) was dissolved in peanut oil at a concentration of 10 mg/ml. To induce Cre activity, pregnant mice were given either IP injection or oral gavage of tamoxifen at a dose and at time points described in the text. A 27-gauge needle was used for IP injection; a 22-gauge feeding needle was used for oral gavage. The morning in which a vaginal plug was detected was designated as E0.5.
Staining for detecting of β-galactosidase activity was performed as previously described (Frank et al.,2007).
Whole-Mount In Situ Hybridization
Embryos were harvested, fixed, and hybridized with a digoxigenin-labeled antisense RNA probe to Foxa2 (Sasaki and Hogan,1993) using a standard protocol (Park et al.,2006). Control and mutant embryos were processed in the same vial throughout. Stained embryos were post-fixed in 4% paraformaldehyde and cleared in 80% glycerol. Embryos were then photographed under a LEICA MZ125 stereomicroscope using a Canon Powershot A95 digital camera.
We thank Michael Reth for the merCremer;neor cassette, Amy Talbot and Ranajeet S Saund for technical assistance, and Kirk Thomas and Debbie Frank for critical reading of the manuscript. J.F.M. and A.M.M were funded by the NIH.