Efficient generation of endogenous protein reporters for mouse development

Mouse models are essential tools for research to uncover human disease mechanisms. To produce new mouse lines, embryos are collected and genetically manipulated during the preimplantation stage, and are then transferred to surrogate mothers for gestation. Thus, preimplantation embryos are the starting point for many studies. Within preimplantation, the blastocyst stage of development is also an alluring model in its own right. This is in part because the blastocyst provides technical advantages, including optical transparency, the capacity to develop ex vivo in a cell culture incubator, and the ease of collecting dozens of embryos at a time. These properties have enabled discovery of the molecular mechanisms of the first steps in mammalian development. Moreover, embryonic stem cells (ESCs) are derived from blastocysts, providing additional models for basic and applied research. Thus, technological advances using the blastocyst can impact broader areas of biomedical research.

One powerful approach to elucidating the molecular mechanisms of development and disease has been live imaging of fluorescent reporters in vivo, which enables time-resolved analysis of gene expression at the cellular level (Nowotschin and Hadjantonakis, 2014). Live imaging of gene expression in vivo is often achieved by knocking in genes encoding green fluorescent protein (GFP) and other fluorescent proteins downstream of gene promoters, to create gene reporters. This approach requires establishing and breeding new mouse lines. An alternative method of protein detection is to use antibodies to localize endogenous proteins, for example by immunofluorescence. However, immunofluorescence does not allow visualization of dynamic processes. Moreover, identification of reliable and specific antibodies can also be time intensive and, for some antigens, may not exist.

Our goal was to help overcome some of these challenges by developing an alternative, streamlined pipeline for the detection or screening of endogenous proteins in vivo. We focus on preimplantation mouse embryos, and present two complementary approaches to enhance the efficiency of detecting endogenous proteins in vivo. We provide guidelines for implementation of these approaches in broader experimental settings.

Like GFP, the yellow-green, monomeric fluorescent protein mNeonGreen (mNG), derived originally from the marine invertebrate Branchiostoma lanceolatum, is an 11-stranded beta-barrel, but is up to three times brighter than GFP (Shaner et al., 2013). The mNG derivative, mNG2, can be split into two separate coding units, mNG2(Δ11), which lacks the 11th beta-strand, and mNG2(11), which is the 11th beta-strand (Feng et al., 2017). Individually, the two resulting proteins lack appreciable fluorescence. However, when the larger protein mNG2(Δ11) is complemented by the 16-amino acid mNG2(11), fluorescence is reconstituted (Fig. 1A), and the two proteins are capable of self-assembly through non-covalent intermolecular interactions (Cabantous et al., 2005).

We sought to make use of this fluorescence complementation strategy to evaluate localization of endogenous proteins in mouse embryos because we reasoned that tagging endogenous proteins with the smaller, 16-amino acid mNG2(11) coding region would be more efficient than knocking in the full-length gene encoding the full-length, 236-amino acid fluorescent protein. Then, to provide the complementary protein, we aimed to establish a mouse line capable of constitutive expression of mNG2(Δ11). Our goal was to introduce an expression construct including cytomegalovirus enhancer, chicken beta-actin promoter, rabbit beta-globin splice acceptor (CAG) sequences and the mNG2(Δ11) coding region into the Rosa26 (R26) locus by homologous recombination (Fig. 1B), which would enable constitutive, ubiquitous expression of mNG2(Δ11) throughout mouse tissues and development (Friedrich and Soriano, 1991). However, prior to attempting knock-in in mouse zygotes, we first established an R26-mNG2(Δ11) ESC line using a CRISPR/Cas9-mediated knock-in strategy (Chu et al., 2016) (see Materials and Methods). These R26-mNG2(Δ11) ESCs provided a renewable source of positive control genomic DNA for subsequent experiments.

To produce a mouse line capable of expressing mNG2(Δ11), we subsequently introduced the mNG2(Δ11) expression construct into the R26 locus in zygotes, following the strategy we had used in ESCs. Injected zygotes were transferred to recipient females, allowed to gestate, and then founder mice carrying mNG2(Δ11) were identified by PCR genotyping (Fig. 1C,D) and genomic sequencing. A single founder mouse was then expanded and bred to homozygosity to establish R26-mNG2(Δ11)/R26-mNG2(Δ11) mice. In principle, providing mNG2(11) in trans to R26-mNG2(Δ11) would lead to reconstitution of the fluorescent protein. For simplicity, we called this the GOGREEN system.

As an initial evaluation of the GOGREEN system, our first test was to determine whether we could detect fluorescence complementation in embryos by epifluorescence microscopy (Fig. 2A). For this test, we generated mRNA encoding mNG2(11)-tagged clathrin, light polypeptide (Clta). For negative controls, mRNA encoding either mNG2(Δ11) or mNG2(11)-Clta were injected individually into wild-type embryos (Fig. 2B). For a positive control, wild-type zygotes were co-injected with mRNAs encoding both R26-mNG2(Δ11) and mNG2(11)-Clta, and these exhibited greatly elevated fluorescence over both negative controls. Finally, R26-mNG2(Δ11)/+ zygotes were injected with mRNA encoding mNG2(11)-Clta, which led to elevated fluorescence at the blastocyst stage, demonstrating functionality of the GOGREEN system in vivo using epifluorescence and an exogenous mNG2(11)-tagged protein.

We next aimed to evaluate the performance of the GOGREEN system when mNG2(11) was endogenously expressed from several genomic loci. Our goal was to derive R26-mNG2(Δ11)/+ zygotes and, in these, perform CRISPR/Cas9-mediated knock-in of mNG2(11) in frame with proteins of interest (Fig. 3A) to produce mNG2(11) fusion proteins capable of complementing mNG2(Δ11) and reporting endogenous protein patterns.

To achieve in-frame mNG2(11) knock-in, we designed targeting constructs encoding the 16-amino acid mNG2(11), plus a three-amino acid linker, flanked by genomic locus-specific homology arms of 30 nucleotides each (Fig. 3B). Resulting targeting constructs ranged from 117 to 120 nucleotides in length, permitting their synthesis as a single-stranded oligodeoxyribonucleotide (ssODN) by a commercial vendor (see Materials and Methods). For our first knock-in attempts, we targeted cytoskeletal proteins, including intermediate filaments and beta-actin, because their subcellular localizations in mouse preimplantation have long been known (Chisholm and Houliston, 1987; Coonen et al., 1993; Reima and Lehtonen, 1985).

We designed CRISPR reagents to knock mNG2(11) in-frame with keratins 8 and 18 (Krt8 and Krt18), as well as actin, beta (Actb). Following injection of the knock-in mixture into R26-mNG2(Δ11)/+ zygotes, embryos were cultured to the blastocyst stage, and then imaged by confocal microscopy. For each knock-in, we observed the very unique fluorescent meshwork of cortical filamentous proteins expected, in accordance with published observations (Fig. 3C) (Coonen et al., 1993; Lim et al., 2020; Ralston and Rossant, 2008; Reima and Lehtonen, 1985). These observations are indicative of faithful protein reporting. Individual embryos were then harvested, and gene targeting was evaluated by PCR (Fig. 3D) and sequencing. In all cases, monoallelic targeting was highly efficient (Fig. 3E). These observations demonstrate the utility of the GOGREEN system for efficiently reporting the localization of endogenous proteins in vivo. Given that the dynamics of cytoskeletal protein localization and turnover during preimplantation development are actively studied (Anani et al., 2014; Schwarz et al., 2015; Zenker et al., 2018), the GOGREEN system could provide new tools because these proteins are usually visualized either in fixed embryos or by injection of mRNAs encoding tagged proteins, both of which could introduce unwanted artifacts.

Thus far, we had evaluated the ability of the GOGREEN system to report endogenous cytoplasmic proteins in vivo. However, we were uncertain whether the GOGREEN system could effectively report the dynamics of endogenous nuclear protein expression, owing to the possibility that the two components of the GOGREEN system might end up separated by the nuclear membrane.

To investigate the performance of the GOGREEN system in visualizing nuclear proteins in vivo, we evaluated fluorescence in embryos after targeting the genes nucleophosmin (Npm1) and NOP58 ribonucleoprotein (Nop58), which both encode nucleolar proteins. As for previous experiments, we targeted mNG2(11) in frame with target genes in the R26-mNG2(Δ11)/+ genetic background. Remarkably, we were able to detect fluorescence within the nuclear compartment (Fig. 4A) in embryonic cells following mNG2(11) knock-in (Fig. 4B,C). Npm1 fluorescence recapitulated the pattern reported by immunofluorescence (Vogt et al., 2012), while the observed Nop58 pattern is novel. These observations indicate that the nuclear envelope does not necessarily present a barrier to fluorescence complementation, in spite of the fact that the GOGREEN components lack nuclear localization sequences.

Having observed that the GOGREEN system can detect both cytoplasmic and nuclear proteins, we next tested its performance in reporting transcription factor localization, because there is great interest in imaging transcription factor dynamics in living preimplantation embryos (Gu et al., 2018; McDole et al., 2011; Posfai et al., 2017; Saiz et al., 2013, 2015). We next evaluated CDX2, YAP1, GATA6 and NANOG, four transcription factors with essential activities during preimplantation development (Frum et al., 2018; Mitsui et al., 2003; Schrode et al., 2014; Strumpf et al., 2005). However, we were unable to detect appreciable fluorescent signal in embryos of any of these four mNG2(11) knock-ins in the mNG2(Δ11) background (Fig. 4D). We therefore developed a second and alternative knock-in tagging strategy for detecting endogenous transcription factors.

We selected the V5 epitope, a 14-amino acid protein derived from the simian virus 5 (SV5) paramyxovirus because its small size promised high knock-in efficiency and because of the existence of low background, commercially available, monoclonal anti-V5 antibody that could be used for immunofluorescent detection of V5-tagged proteins in embryos. We then designed V5-encoding targeting constructs for generating in-frame V5 fusion proteins (Fig. 5A,B).

We targeted the V5 tag in frame with key transcription factors in zygotes, and then observed immunofluorescence patterns in blastocysts by confocal microscopy. We were able to detect V5 signals that were clear and specific after targeting the nuclear factors such as GATA3, CTCF and NANOG (Fig. 5C-E). Importantly, the patterns of V5-tagged GATA3, NANOG and CTCF recapitulated their reported expression patterns in blastocysts (Home et al., 2009; Marcho et al., 2015; Ralston et al., 2010; Strumpf et al., 2005). In a parallel set of experiments, we harvested embryos prior to blastocyst stage and then co-stained these with anti-V5 and anti-NANOG antibodies. In these embryos, we detected NANOG expression in every V5-expressing cell (Fig. S1), confirming the utility of the V5 knock-in approach for faithfully reporting gene expression, even at preimplantation stages prior to blastocyst. These observations highlight the utility of the V5-tagging system to evaluate the endogenous expression patterns of known or novel transcription factors.

Having observed that V5 outperformed the GOGREEN system, in terms of transcription factor detection, we hypothesized that protein abundance could be the limiting factor for detection using the GOGREEN system. Consistent with this hypothesis, we were able to detect the transcription factor CDX2 using GOGREEN when Cdx2-mNG2(11) was overexpressed by mRNA injection (Fig. 6A,B). Finally, we evaluated the abundance of transcripts encoding proteins evaluated in this study, as measured by RNA sequencing (RNA-seq) of individual blastocysts (Aksoy et al., 2013). Remarkably, transcript abundance predicted protein detectability using the GOGREEN system (Fig. 6C). Moreover, proteins of extremely low abundance could still be detected using the V5 system. This analysis therefore provides a guideline for informing subsequent experimental design and for selecting the optimal protein tagging approach. Ultimately, the GOGREEN and V5 systems together enable detection of endogenous proteins across the range of protein expression levels, facilitating multiple downstream applications and opening doors for new discoveries.

Split GFP and V5 epitope tagging have been used for protein detection in cell lines and in some animal models (Hefel and Smolikove, 2019; Kamiyama et al., 2016; Kim et al., 2012; Leonetti et al., 2016; To et al., 2016; Yamagata and Sanes, 2012; Yang et al., 2013), but their use as knock-in mouse reporters has not been systematically compared across diverse genomic loci. Here, we presented a systematic comparison of their performance, sensitivity and efficiency of endogenous protein reporting. We note that both approaches are similarly efficient, averaging ∼60% knock-in efficiency across more than a dozen loci tested. This rate is much higher than targeting full-length fluorescent proteins by zygote injection. For example, we observed a 6% knock-in efficiency at the R26 locus with the nearly full-length fluorescent protein mNG2(Δ11). In fact, we observed upwards of 75-100% knock-in efficiency for multiple loci, which exceeds allele inheritance rates in most mating strategies. Moreover, the relatively short length of the ssODN enables higher efficiency targeting and ease of synthesis, bypassing traditional molecular cloning methods required for producing longer donors. Finally, tagging with GOGREEN and V5 enables efficient detection of endogenous proteins, thereby circumventing artifacts caused by imaging fluorescently tagged, overexpressed proteins.

We note opportunities for applying biochemical and molecular techniques in vivo. V5 is commonly used for purifying proteins from cells and tissues for the downstream identification of protein or nucleotide interactions, including immunoprecipitation-western blotting or mass spectrometry, chromatin-immunoprecipitation or ribonucleotide pulldown and sequencing (ChIP-seq, RIP-seq), or cleavage under targets and release using nuclease (CUT&RUN) (Hainer et al., 2019; Skene and Henikoff, 2017). We therefore envision that the approaches described here could be used to generate stable mouse lines that enable anti-V5 antibody-mediated discovery of protein localization patterns, protein- and RNA-binding partners and DNA-binding sites throughout the genome.

Both GOGREEN and V5 systems present exciting opportunities for biological investigation outside of preimplantation mouse development as well. For example, V5 or mNG2(11) knock-in embryos could be transferred to recipient females to allow for postimplantation development so that protein localization can be evaluated in later developmental processes or in adult tissues and organs. Additionally, both the GOGREEN and V5 systems could be adaptable to viral transduction (Yoon et al., 2018), which could extend applications to adult organs and tissues. Our studies thus provide guidelines, molecular reagents and genotyping assays to enable these applications.

In considering endogenous protein tagging applications, we identify several key considerations. First, care should be given to the design of the tagged protein, and whether the location and nature of the tag interfere with protein function. Validation for protein function and localization can be confirmed using appropriate strategies, including mouse genetics and, if possible, by confirming protein localization by immunofluorescence. Second, guide RNA (gRNA) design should follow best practices so as to minimize the chance of on/off-target indel alleles; targeting protein C-termini may help avoid unwanted phenotypes caused by frame-shift mutations. Third, the genotyping strategy should confirm that the tag has been knocked in in frame with the target protein at the sequence level. Related to this, strategies for identifying random ssODN insertions should be considered (Lanza et al., 2018). Finally, if microinjection is to be used as the delivery method, consultation with institutional transgenic facility with proper technical expertise should be sought, when available, to ensure optimal experimental design.

Finally, both the GOGREEN and V5 systems could also be used in embryos from species such as humans or other primates, for which breeding to establish knock-in lines is either inappropriate or impractical. There would be additional advantages to applying either system to emerging mammalian models, such as marsupials, for which protein-specific antibodies have not yet been developed. For live imaging, mNG2(Δ11) could be provided by mRNA injection, while mNG2(11) would be knocked in frame into genes of interest. If fixed imaging of low abundance proteins is preferred, then V5 could be knocked in. Either system promises new opportunities for the discovery of developmental principles in mouse as well as understudied mammalian species.

All animal research was conducted in accordance with the guidelines and approval of the Michigan State University (MSU) Institutional Animal Care and Use Committee. Most experiments were performed using male or female CD-1 mice, at least 6-8 weeks of age, maintained on a 12-h day/night cycle with food and water ad libitum.

The pR26-CAG-mNG2(Δ11) targeting vector was cloned by insertion of a synthesized dsDNA fragment encoding mNG2(Δ11) (Table S4) into the previously published vector pR26-AsisI/MluI (Addgene #74286) (Chu et al., 2016) via restriction/ligation with AsisI and MluI. After cloning, the Lox-Stop-Lox site was removed by exposure to recombinant Cre recombinase (NEB), using the NEB standard protocol. The in vitro transcription plasmids for Clta-mNG2(11), mNG2(Δ11) and Cdx2-mNG2(11) were cloned by inserting a synthesized dsDNA fragment (Table S4) containing the respective coding sequence into a pcDNA3.1-poly(A)83 vector (Yamagata et al., 2005) downstream of the T7 promoter via restriction ligation with HindIII and NotI. pX459-sgR26-1 was generated by inserting the gRNA sequence targeting R26 (Table S3) into pSpCas9(BB)-2A-Puro (PX459) V2.0 (Addgene #62988) via restriction ligation with BbsI.

In vitro transcription (IVT) was performed using a T7 mMessage mMachine kit (Life Technologies). Each IVT construct was digested with XbaI, followed by ethanol precipitation, and was then used in the IVT reaction as per the manufacturer's instructions. Resulting mRNA was purified using a MEGAclear kit (Ambion). mRNA quantity and quality were assessed by a Nanodrop spectrophotometer and by agarose gel, respectively.

Target-specific crispr RNA (crRNA) and non-variable trans-activating crispr RNA (tracrRNA) obtained from Integrated DNA Technologies (IDT) were each suspended in injection buffer (1 mM Tris-HCl, pH 7.5; 0.1 mM EDTA), mixed at a 1:1 molar ratio and annealed in a thermal cycler by ramping down from 95°C to 25°C at 0.1°C/s. The annealed RNAs were then mixed with recombinant Cas9 protein at a 5:1 RNA:Cas9 molar ratio and allowed to form ribonucleoproteins (RNPs) for 15 min at room temperature. RNPs were mixed with donor ssODN synthesized as Ultramers from IDT (Table S2) and diluted to working concentrations (Table S5) by addition of injection buffer. For mRNA injections, each mRNA was diluted in injection buffer to 350 ng/µl and injected either into one pronucleus of the mouse zygote or into one blastomere of the two-cell mouse embryo. Injection mixes and mRNAs were aliquoted and stored at −80°C, avoiding freeze/thaw cycles.

Zygotes were harvested from naturally mated pregnant mice on the day that copulatory plugs were detected. Oviducts were flushed with M2 medium (Millipore Sigma), and then injection mix was delivered into one pronucleus or the nucleus of one blastomere via microinjection (Nagy et al., 2003). Injected zygotes were cultured in KSOM+amino acids (AA) (Millipore Sigma) for up to 5 days before being fixed or imaged live. Only embryos that survived injection and appeared to have cavitated or to be attempting cavitation were included in the analysis.

The R26-mNG2(Δ11) mouse line was generated by zygote microinjection at the MSU Transgenic and Genome Editing Facility. R26 sgRNA was synthesized by IVT of a PCR-amplified region of pX459-sgR26-1 using the Life Technologies MEGAshortscript T7 kit. Transcripts were subsequently purified using the MEGAclear kit. A mixture containing 5 ng/µl circular pR26-CAG-mNG(Δ11) and 125 ng/µl Cas9 RNP complexed with sgR26 gRNA was injected into one pronucleus of C57BL/6J mouse zygotes. Zygotes were then transferred to CD-1 recipient mice. After birth, tail tips were screened by PCR for successful integration of the R26-CAG-mNG(Δ11) allele.

V5 was detected by mouse anti-V5 antibody (Thermo Fisher Scientific, R96025). Embryos were fixed with 4% formaldehyde (Polysciences) for 10 min, permeabilized with 0.5% Triton X-100 (Millipore Sigma) for 30 min, and then blocked in 10% fetal bovine serum (Hyclone) with 0.1% Triton X-100 for 1 h at room temperature. Embryos were then incubated in anti-V5 antibody at a dilution of 1:400 in blocking solution at 4°C overnight. The next day, embryos were stained with goat anti-mouse Alexa488 (Invitrogen, A-11030) at a 1:400 dilution in blocking solution for 1 h at room temperature. Embryos were then stained for 10 min at room temperature in 50 µM Hoechst nucleic acid stain (Thermo Fisher Scientific) or DRAQ5 (Cell Signaling Technology, 4084S; 1:400 dilution). Rabbit-anti-NANOG (Reprocell, RCAB002P-F) was used at 1:400 dilution, with Cy3-conjugated donkey-anti-rabbit IgG (Jackson ImmunoResearch, 711-165-152) at 1:400 dilution. Split mNG2 embryos were imaged either fixed or live after Hoechst staining. Imaging was performed using an Olympus FluoView FV1000 Confocal Laser Scanning Microscope system with a 20× UPlanFLN objective (0.5 NA) and 5× digital zoom or with a 60× PlanApoN oil (NA 1.42) objective. For each embryo, z-stacks were collected with 3 µm intervals between optical sections. Optical sections are displayed as an intensity projection over the z-axis. Figures were prepared using FIJI, Adobe Photoshop and Adobe Illustrator.

To genotype embryos, genomic DNA was extracted from single blastocysts by placing each blastocyst in a microtube containing 4.4 µl extraction buffer (REDExtract-N-Amp Tissue PCR Kit, Millipore Sigma) mixed with 1.1 µl tissue prep buffer, and then incubating tubes at 56°C for 30 min, 24°C for 5 min and 95°C for 5 min. After incubation, 5 µl neutralization buffer was added to each tube. In subsequent reactions, 1 µl of embryo extract was used as a PCR template, and locus-specific primers (Table S1). To genotype adult mice, genomic DNA was extracted from ear punch biopsies using a Wizard SV Genomic DNA Purification System (Promega), and PCR was performed using Herculase II Polymerase (Agilent).

To confirm the identity of select PCR products, the products were directly cloned into pCR2.1 TOPO using an Invitrogen TOPO TA Cloning Kit (Invitrogen). Plasmids containing the PCR product were prepped with the Spin Miniprep Kit (Qiagen) and then sequenced by RTSF Sanger method at the Genomics Core at MSU.

R1/E ESCs (American Type Culture Collection) were cultured on CF-1 feeder mouse embryonic fibroblasts (MEFs; Applied Stem Cell) in ESC medium [Dulbecco's modified Eagle medium (DMEM) supplemented with 1000 U/ml leukemia inhibitory factor (Millipore Sigma), 15% (v/v) fetal bovine serum (Hyclone), 2 mM L-Glutamax (Thermo Fisher Scientific), 0.1 mM beta-mercaptoethanol (Millipore Sigma), 0.1 mM minimum essential medium (MEM) non-essential amino acids (Millipore Sigma), 1 mM sodium pyruvate (Millipore Sigma) and 1% (v/v) penicillin/streptomycin (Gibco)]. Passage 13 R1/E ESCs were cultured to ∼70% confluence in a 10 cm dish, and then electroporated with pX459-sgR26-1 and pR26-CAG-mNG2(Δ11) as follows: pelleted cells were resuspended in 800 µl Embryo-Max Electroporation Buffer (Millipore Sigma) containing 20 µg of each plasmid, and cells were then electroporated in a 0.4 cm electrode gap electroporation cuvette (Bio-Rad) using a Bio-Rad Gene Pulser XCell electoporator (250 V, 500 µF, infinite Ω). Subsequently, 400 µl electroporated cells were diluted in 10 ml ESC medium and then plated on a 10 cm dish on puromycin-resistant DR4 feeder MEFs (Applied Stem Cell). After 24 h, selection was started with ESC medium containing 1.25 µg/ml puromycin (Gibco). After 12 days, colonies were picked into 96-well plates and expanded over several more passages. Cell lines were genotyped by PCR using R26F3 and SAR primers to detect insertion of mNG2(Δ11) in the R26 locus (Table S1).

We thank Dr Kristin Parent for comments on the manuscript; current and former members of the Ralston Laboratory, especially Dr Michael Halbisen and Dr Tristan Frum, for helpful discussions and training; Axel Schmitter for assistance in the laboratory; and Drs Elena Demeriva and Huirong Xie at the MSU Transgenic and Genome Editing Facility for the R26 mouse knock-in line.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.197418