Embryogenesis is a dynamic process involving cell division, cell movement, cell death and morphological changes of cells and tissues. An increasing number of fluorescent proteins (FPs) have been developed, and a set of protein domains that mediate subcellular localization (SCLDs) have been identified. Various fusion fluorescent proteins (FFPs) in which FPs are fused to SCLDs have been developed to monitor cell behaviors during embryogenesis (Cormack et al.,1996; Nowotschin et al.,2009b; Nowotschin and Hadjantonakis,2009; Shaner et al.,2004). Among them, FPs fused at the N-terminus to histone H2B (H2B-FPs) and FPs fused at the C-terminus to a glycosylphosphatidylinositol (GPI) signal sequence (FPs-GPI) have been demonstrated especially useful to highlight nuclei and cell membranes, respectively (Hadjantonakis and Papaioannou,2004; Kanda et al.,1998; Nowotschin et al.,2009a; Srinivas et al.,2001).
Mice in which multiple cell organelles are labeled simultaneously in different colors are useful to monitor cell behaviors in vivo. However, it is laborious to establish and maintain such mice by crosses among different transgenic mouse lines harboring each individual FFP. Several approaches have been made to overcome this problem (Szymczak and Vignali,2005). One such approach is the use of the self-cleavage 2A peptide sequence (de Felipe et al.,2006). Transgenic mice were successfully generated with a Tom-2A-GFP construct in which Myr-TdTomato and H2B-GFP were linked by the 2A sequence, being directed by a ubiquitous CAG promoter (Trichas et al.,2008). Transgenic mice were also generated with a GR (green–red) construct in which H2B-EGFP and mCherry-GPI were linked by the 2A sequence, being directed by a ubiquitous UBC promoter (UBC-GR) and a tissue-specific Oct4 promoter (Oct4-GR) (Stewart et al.,2009). Using a similar construct, RG, in which H2B-mCherry and EGFP-GPI were linked by the 2A sequence (Stewart et al.,2009) (Fig. 1a), the present study planned to take a different approach of expressing the RG construct conditionally. Conditional expression can be accomplished by inserting the RG construct that has a loxP-flanked transcription stop signal located 5′ upstream into a ubiquitously active genome locus by homologous recombination (R-RGallele; Fig. 1b) (Srinivas et al.,2001). We have chosen the Rosa26 (R26) locus as such a locus (Soriano,1999); the intervening sequence used in this study was the neomycin resistant gene (Neo) directed by a PGK promoter with an SV40 polyadenylation signal triply repeated (tpA) (Fig. 1b). We call the reporter mouse line R26R-RG.
The R26 locus has been demonstrated to have ubiquitous transcriptional activity during development (Hohenstein et al.,2008; Kisseberth et al.,1999; Soriano,1999; Srinivas et al.,2001; Zambrowicz et al.,1997). We tested whether this locus has the transcriptional activity sufficient for in vivo live-imaging. Neither heterozygous R26+/R-RG nor homozygous R26R-RG/R-RG (R26R-RG) embryos exhibits any fluorescent signal (data not shown); they are live-born, normal and fertile without any apparent defects beyond one year. The expression of the RG construct will generally be made by the cross of homozygous R26R-RG mice with the mouse expressing Cre in the specificity of one's choice (RG allele; Fig. 1b). Consequently, the live-imaging will be made in the heterozygous R26-RG state. The question of whether the heterozygous state expresses the RG at sufficient intensity was examined in E6.0-E7.5 heterozygous R26-RG embryos; the primitive streak is formed at E6.5, and E7.5 is the stage when active cell migration, proliferation and morphological changes occur. For this examination, heterozygous R26-RG mice that lack the intervening sequence in front of RG in R26R-RG was established by the cross of homozygous R26R-RG mice with EIIa-Cre mice that express Cre ubiquitously including germ cells (Lakso et al.,1996); homozygous R26-RG mice were obtained by their brother-sister mating. Not only heterozygous but also homozygous R26-RG mice are live-born, fertile, and apparently normal beyond one year. The E7.5 heterozygous R26-RG embryos exhibited red and green fluorescence in nuclei and membranes, respectively, in every cell of the body at sufficient resolution and intensity (Fig. 1c).
Next, heterozygous and homozygous R26-RG embryos were cultured, and time-lapse images were examined. This demonstrated that the cells of the visceral endoderm changed drastically in shape and positions in both heterozygous and homozygous E6.5 R26-RG embryos (Supporting Information movie 1), chromosome segregation during cell division was also clearly observed. In epiblast cells it is known that nuclei locate basally during interphase, whereas they translocate apically during mitosis (Gardner and Cockroft,1998). The cells locate deeper in the embryo, and the imaging was not distinct enough. Nevertheless, the live-imaging demonstrated nuclear movements in E6.0 epiblast cells from the apical side to the basal side and vice versa, and the occurrence of cell divisions near the apical side (Fig 2, Supporting Information movie 2). The imaging also demonstrated the shape change of epiblast cells from columnar to round, accompanying the nuclear movement. The change of cell shape and chromosome segregation during cell division was also clearly observed in ectoplacental cone cells (Fig. 3, Supporting Information movie 3). During the time-lapse observation for 24 h the embryos were as healthy and normal as embryos not illuminated.
Live images were also obtained with heterozygous R26-RG E5.5 embryos at a similar resolution and intensity (data not shown); heterozygous expression is sufficient for live-imaging in postimplantation embryos. However, the heterozygous expression was not sufficient in preimplantation embryos. In mouse, zygotic gene expression occurs from the 2-cell stage, and maternal expression has a significant contribution to the gene expression in preimplantation embryos. When time-lapse observations were made from the 2-cell stage to blastocyst-stage, heterozygous R26-RG embryos that inherited RG paternally exhibited no fluorescence at 2- and 8- cell stages and very weak fluorescence at the blastocyst stage (data not shown). The heterozygous R26-RG embryos that inherited RG maternally exhibited weak fluorescence at 2- and 8-cell stages; the fluorescence increased significantly at the blastocyst stage but was not sufficient enough for time-lapse imaging (data not shown). In homozygous R26-RG embryos, red and green fluorescence was observed in all cells from the 2-cell to blastocyst stage (see Fig. 4); under the conditions examined the embryos developed normally and hatched from the zona pellucida (Fig. 4, Supporting Information movie 4). In these embryos red H2B-mCherry fluorescence was intensively detected in nuclei as expected. However, in contrast to postimplantation stages, EGFP signals were prominent not only in plasma membranes but also in cytoplasm at the blastocyst stage; the cytoplasmic green signals became fewer and fewer at earlier stages. In addition, green EGFP was also apparent in the nucleus at the 2-cell to 8-cell stage. Concomitantly, the intensity of green fluorescence in cell membranes was not very high in the preimplantation embryos.
Tissue-specific RG expression was next examined in Sox9-Cre; R26R-RG embryos that were obtained by breeding R26R-RG females with Sox9-Cre knock-in males (Akiyama et al.,2005). Sox9 is expressed in a variety of fetal tissues, including the developing skeleton and Sertoli cells of the testis (Kobayashi et al.,2005). Sox9-Cre; R26R-RG embryos at E13.5 had mCherry and EGFP fluorescence in the craniofacial and limb/trunk skeleton-forming regions (Fig. 5a–c). In E13.5 testes, mCherry and EGFP signals highlighted the testicular cords (Fig. 5d–f) but not mesonephros (data not shown). Sox9-Cre mediated R26R-RG reporter activity was not detected in E13.5 ovaries (data not shown). In Sox9-Cre; R26R-RG testes, Sertoli cells but not interstitial cell types (e.g., Leydig cells) were labeled by both nuclear mCherry and membranous EGFP signals (Fig. 5g–i). This pattern of reporter activity in the developing testes correlates with Sox9 expression in fetal Sertoli cells; Sox9 is not expressed in interstitial cell types (Kobayashi et al.,2005). These data suggest that the R26R-RG reporter mouse can be used for highlighting nuclear and cell membrane morphologies in specific cell types, using tissue-specific Cre transgenic mice.
Mouse embryonic fibroblast (MEF) cells were also prepared from E14.5 embryos to assess the suitability of time-lapse imaging in cells cultured from heterozygous R26-RG embryos (Fig. 6, Supporting Information movie 5). Red mCherry signals were intense in nuclei and no green EGFP signals were observed in nuclei. However, EGFP signals were weak in the cell periphery; this is probably because MEFs are very flat. EGFP signals were brighter at lamellipodia and during mitosis when cells became thicker (see Fig. 6).
We propose to generate a bank of reporter lines not by random transgenesis, but conditionally using an endogenous locus ubiquitously active and of which loss does not affect the fertility or viability of mouse. In reporter mice generated by random transgenesis (pronuclear DNA injection), expression patterns in each tissue vary between mouse lines due to differences in copy number and positional effects of integration sites. The screening of many transgenic lines, which is done to identify the reporter line suitable for the live-imaging, is a laborious and time-consuming process and expensive. Furthermore, reporter mice that exhibit high fluorescent expression are frequently infertile or not viable (Stewart et al.,2009). Cell behaviors frequently need to be examined by analysis with multiple reporter lines with FPs fused with SLCDs, but it is too difficult to establish several different reporter lines by random mutagenesis. Conditional reporter lines invite researchers to generate a Cre line suitable for his/her interest. A tissue-specific reporter line established by random mutagenesis can be used just for the live-imaging of the tissue with that reporter. In contrast, a tissue-specific Cre line can be used not only to generate a series of reporter mice for live cell imaging by the cross with a series of conditional reporter lines, but also to conditionally mutate a variety of genes. An increasing number of Cre mouse lines is indeed accumulating, including inducible Cre lines such as Cre combined with an estrogen responsible element (Danielian et al.,1998; Nagy,2000).
Several studies have suggested that the Rosa26 locus can serve for the production of conditional reporter lines, and this study also supports this. The generation of mutant mice is routinely accomplished at this locus since the frequency of homologous recombination is relatively high. The parent conditional R26R-FFP mice should be normal, fertile and established as the homozygous line, as demonstrated by the homozygous R26R-RG mouse line in this study. We have already established 17 mouse lines that express a series of FFPs (Abe et al., this issue), including the R26R-RG line established in this study. The R26R-RG conditional reporter mouse line yielded the R26-RG mouse line in which the nucleus and cell membranes are marked by mCherry and EGFP, respectively, ubiquitously. R26-RG mice can be maintained in the homozygous state, and should serve as valuable mouse lines for live-imaging.
At the same time, the fertility in homozygous R26-RG mice indicates that the expression level of FFPs is not high enough to be toxic at this locus. Indeed, the intensity of fluorescent signals was lower than that in UBC-HS-GR transgenic mice which were infertile at a heterozygous state (Stewart et al.,2009); the intensity might also be lower than that in Tom-2A-GFP transgenic mice by Trichas et al. (2008) which is fertile. Using R26-RG embryos, we were able to observe the nuclear movement in the epiblast cells, located deeply within the embryos, in live-imaging with a conventional confocal microscope, but the quality was not high enough. The next step is to improve the quality by the conditional approach. One choice is to change the combination of the FFP gene. GPI-anchored proteins are usually sorted into rafts and unevenly distributed in membranes (Kondoh et al.,1999). We have surveyed a variety of fluorescent genes in combination with a series of SCLDs at Rosa26 locus for their activity in live-imaging (Abe et al., this issue). In detecting membrane, Lyn-Venus is much brighter than EGFP-GPI; Myr-TdTomato might also be brighter (Trichas et al.,2008). Higher expression may also be established by enhancing the transcriptional activity of the Rosa26 locus; for example, by placing ubiquitously strong enhancer in front of the stop codon flanked by loxP. The insertion of FFP genes in multiple copies might also be useful. It may also be accomplished by the identification of another gene locus which has more intensive and ubiquitous expression activity than Rosa26 and of which loss is not lethal or infertile. At the same time, we expect the development of less toxic FPs and of a less toxic and deeply penetrating conventional illuminating system, together with a more sensitive detection system.
The bicistronic expression of H2B-mCherry and EGFP-GPI at the nucleus and plasma membranes by 2A peptide in this manuscript is largely consistent with those shown by Stewart et al. (2009) (the expression of H2B-EGFP and mCherry-GPI) and by Trichas et al. (2008) (the expression of H2B-GFP and Myr-TdTomato), especially at postimplantation stages. In our study EGFP-GPI signal was prominent in cytoplasm at the blastocyst stage; Trichas et al. (2008) also observed Myr-TdTomato signal in cytoplasm, as did Stewart et al. (2009) mCherry-GPI signal in cytoplasm of several adult tissues. However, Trichas et al. (2008) noted that they never detected Myr-TdTomato in the nucleus and H2B-GFP never in the plasma membrane or cytoplasm in their Tom-2A-GFP transgenic mice. In this study, green EGFP-GPI fluorescence was found in the nucleus, in addition to the expected expression at cell membranes in the 2-cell to 8-cell stage RG embryos. Furthermore, some EGFP-GPI fluorescence was also observed in the nucleus of E6.5 ectoplacental cone cells (see Fig. 3). We initially thought that the nuclear EGFP-GPI may be expressed as the fusion protein with H2B-mCherry, and that its localization may be directed by the H2B nuclear localizing signal. EGFP-GPI expression pattern in the nucleus was indeed very similar to the H2B-mCherry expression pattern; even at mitotic phase when nuclear membrane disappears, their expression is retained, probably in chromatin. However, H2B-mCherry was not apparent in plasma membranes when EGFP-GPI fluorescence was found in the nucleus; this is contrary to the report on the fusion protein localization by Trichas et al. (2008) with a mutant 2A.
Another possibility is that the “green” fluorescence in the nucleus is actually red mCherry fluorescence excited by EGFP excitation light at 488 nm. The fluorescence was detected with a monochrome camera through a triple band filter (transmission band: 495–555 nm, 575–635 nm, and 665–775 nm) for both EGFP and mCherry emission signals, exciting either at 488 nm for EGFP or 561 nm for mCherry. The fluorescence detected through the filter after the 488 nm excitation was regarded as EGFP emission (maximum emission: 507 nm) and colored green, and that after 561 nm excitation as mCherry emission (maximum emission: 610 nm) and colored red; images were given as their merged views. The possibility that mCherry fluoresced by 488 nm excitation was examined at the two cell stage. A R26-RG embryo was excited at 488 nm, and the emission was scanned at 10 nm width over 489-708 nm (Supporting Information Fig. 1). The EGFP fluorescence at 499–518 nm was found at plasma membranes and cytoplasm, but not at the nucleus. Moreover, with 488-nm excitation there was a weak fluorescence in the nucleus at 609–618 nm which must be mCherry emission. Therefore, we consider that the “green” fluorescence in the nucleus is actually “red” mCherry fluorescence by 488-nm excitation, and that this was relatively more enhanced by lower EGFP expression at plasma membranes at 2- and 8-cell stage than that at postimplantation stages.
It is not clear why EGFP green fluorescence in plasma membranes was weaker than mCherry red fluorescence in the nucleus at the preimplantation stages. It is known that even when the self-cleavage occurs at nearly 100% frequency with a 2A sequence in a cell free system, the self-cleavage sometimes does not take place at 100% frequency in cultured cells with the same 2A sequence; it occurs at some frequency where the C-terminal side of the 2A is less efficiently translated (de Felipe et al.,2006). The frequency is known to be affected by protein structures at both sides of 2A, especially at the N-terminal side (Szymczak et al.,2004), and this should be taken into consideration in designing the 2A-mediated dual reporters in future studies. It may be better to get higher signals in plasma membranes by placing the reporter gene that marks plasma membranes at the N terminal side of 2A. It is also possible that there are factors that affect the self-cleavage activity of 2A, and that the expression of such factors changes developmentally; translation termination release factors eRF1 and eRF3 are known to play an important role in the function of the 2A peptide (Doronina et al.,2008).
In summary, the R26R-RG mouse line was established in which H2B-mCherry and EGFP-GPI linked by 2A sequence were conditionally inserted into the Rosa26 locus. The intensity and resolution of mCherry nuclear and EGFP membrane localization were sufficient for live-imaging, and the reporter mouse line should be a useful tool for labeling nuclei and membranes simultaneously in distinct cell populations.