A cell cycle-independent, conditional gene inactivation strategy for differentially tagging wild-type and mutant cells

Here, we describe a novel method based on intronic MiMIC insertions described in Nagarkar-Jaiswal et al. (2015) to perform conditional gene inactivation in Drosophila. Mosaic analysis in Drosophila cannot be easily performed in post-mitotic cells. We therefore, therefore, developed Flip-Flop, a flippase-dependent in vivo cassette-inversion method that marks wild-type cells with the endogenous EGFP-tagged protein, whereas mutant cells are marked with mCherry upon inversion. We document the ease and usefulness of this strategy in differential tagging of wild-type and mutant cells in mosaics. We use this approach to phenotypically characterize the loss of SNF4Aγ, encoding the γ subunit of the AMP Kinase complex. The Flip-Flop method is efficient and reliable, and permits conditional gene inactivation based on both spatial and temporal cues, in a cell cycle-, and developmental stage-independent fashion, creating a platform for systematic screens of gene function in developing and adult flies with unprecedented detail. DOI: http://dx.doi.org/10.7554/eLife.26420.001


Introduction
Functional gene annotation requires, among others, the knowledge of phenotypes observed in mutants of the gene of interest, as well as the expression pattern and subcellular localization of the protein. For Drosophila genes whose loss causes early developmental lethality, characterization of gene function in later stages of the animal's life cycle relies on generating mosaics. Currently, Mosaic Analysis with a Repressible Cell Marker (MARCM) (Lee and Luo, 1999) and CRISPR/Cas9-mediated somatic mutagenesis (Port et al., 2014; are the two major techniques used to study mutations in mosaics. While MARCM has been very successfully used in mitotically active, developing tissues, the technique has two major limitations. First, the technique relies on mitotic recombination and can therefore not be used in non-dividing cells. Second, along with the mutation under study, other mutations in the chromosome distal to the FRT site also become homozygous in mutant cells, requiring 'rescue experiments' of the gene of interest to validate the results (Roegiers et al., 2009). Another strategy is based on CRISPR/Cas9-mediated somatic mutagenesis to generate mosaics (Port et al., 2014;. The limitations of CRISPR/Cas9 are that the identity of the generated lesions for the gene of interest may vary in individual cells in the same animal or tissue. Moreover, the mutant cells are not marked and cannot be distinguished from neighboring wild-type cells. Here, we describe a new flippase (FLP)-dependent method 'Flip-Flop' that offers several advantages over the current techniques for generating mosaics. (1) The method does not rely on cell division and can, therefore, be broadly used for conditional gene inactivation in post-mitotic cells such as neurons.
(2) It allows endogenous tagging of proteins with EGFP, which permits multiple applications, and (3) it simultaneously marks mutant cells with mCherry.

Results and discussion
The Flip-Flop construct and strategy We engineered the 'Flip-Flop' cassette for conditional gene inactivation. This cassette contains two modules that are placed in opposite orientation: a protein-trap (PT) module and a gene-trap (GT) module ( Figure 1A). The PT module carries a splice acceptor (SA), followed by an in-frame EGFP coding sequence, and a splice donor (SD) (Venken et al., 2011). The GT module similarly contains a SA, but is followed by the T2A peptide sequence, an mCherry coding sequence, stop codons in all three reading frames, and an SV40 polyA signal. The PT and GT modules are placed in opposite orientations and are flanked by inverted pairs of canonical FRT and FRT14 sites, forming a flip-excision switch (FLEx) (Schnütgen et al., 2003;. The entire cassette is nested between two inverted attB sites that facilitate Recombination-Mediated Cassette Exchange (RMCE) between the Flip-Flop cassette and a target Minos-Mediated Integration Cassette (MiMIC) that resides in a coding intron of the gene of interest. When integrated in the MiMIC in the PT orientation, Flip-Flop should result in expression of the endogenous protein with an internal EGFP tag. The internal EGFP tagging does not or subtly disrupt the protein's function in 77% of the cases . The cassette can then be converted from a PT to a GT, in vivo, by inverting the cassette's orientation through the expression of FLP that acts on the FLEx switch ( Figure 1B). Following the switch from the PT to the GT orientation, transcription is precociously terminated by the polyA sequence. When this truncated transcript is translated, the T2A site induces a translational skip (Tang et al., 2009), truncating the native protein and re-initiating translation at the mCherry sequence ( Figure 1B). Hence, inversion of the Flip-Flop cassette results in the generation of a truncated protein, which is typically non-functional, and simultaneously marks the cells that are actively eLife digest The instructions needed to build and maintain cells in an organism are encoded in their DNA. There are many different cell types, and each type only needs a small portion of the information found in the DNA to do its job. Hence, only some of the instructions, in the form of genes, need to be active or 'expressed' in any given cell type.
To understand how a gene works, it is necessary to know in which cell the gene is expressed and where in the cell the gene product -normally a protein -is located. Researchers may study a gene by deleting it, which prevents the protein from being made, or by attaching a new instruction into the gene, which generates a fluorescent tag on the protein to determine where and when it is expressed. Until now, it was not possible to selectively inactivate a gene and simultaneously mark both normal cells containing the protein and mutant cells lacking the protein.
Based on an existing tagging approach, Nagarkar-Jaiswal et al. have now developed a method in which normal and mutant cells of fruit flies are marked differently. A gene of interest is tagged with a fluorescent marker called green fluorescent protein (or GFP). The same gene is then inactivated in some of the cells, which are tagged with a red marker called mCherry. Nagarkar-Jaiswal et al. compared normal and mutant cells, and were able to determine how long it takes before the mutant cells become abnormal.
With this new method, the role of numerous genes in any tissue of adult flies can be reassessed. This will allow to investigate what happens when a protein is removed in specific cells in adult flies. A future goal will be to apply this method to other animals that are more closely related to humans, such as mice, to gain a clearer picture of the role of genes in different cell types and how faulty genes may cause disease. The cassette consists of two independent modules (PT and GT), that are oriented in opposite orientations. The PT module contains a splice acceptor (SA), followed by an EGFP tag and a splice donor (SD). The GT module contains a SA sequence, followed by the T2A peptide coding sequence (which will induce a translational skip), the mCherry coding region, stop codons in all three coding frames, and an SV40 polyA transcriptional termination signal. Given the opposite orientation of both Figure 1 continued on next page transcribing the gene with mCherry. While mCherry's expression pattern recapitulates the spatiotemporal expression pattern of the recipient gene, it does not reproduce the endogenous subcellular localization of the protein ( Figure 1B). In summary, expression of FLP will induce a Flip-Flop and produce mCherry-labeled mutant cells in which the gene is inactivated, whereas the surrounding cells are wild-type and express the EGFP-tagged protein.

Flip-Flop-mediated mosaics in mitotically active cells
We created PT insertions for nine genes using available MiMIC insertions: effete (eff), Nedd8, SNF4/ AMP-activated protein kinase gamma subunit (SNF4Ag) Chondrocyte-derived ezrin-like domain containing protein (Cdep), Tripartite motif containing 9 (Trim9), Salt-Induced kinase3 (Sik3), Ankyrin2 (Ank2), Circadian trip (Ctrip) and Ecdysone-induced protein 63E (Eip63E). Seven of the nine generated PT insertions complement the lethality associated with loss of the corresponding gene, suggesting that these internally tagged proteins are biologically functional, consistent with previous observations Venken et al., 2011). The PT insertion in eff  modules, only one of the SA sequence will be active with respect to the recipient gene. The two modules are nested within a pair of FRT and FRT14 inverted repeats, forming a flippase-responsive FLEx switch. Finally, the entire cassette is flanked by two inverted attB sequences that permit phiC31mediated RMCE between the Flip-Flop cassette and pre-existing MiMIC elements. A comparison of the FRT and FRT14 sequence is shown below. The FRT14 sequence varies from the canonical FRT sequence at the residues highlighted in red. (B) Schematic showing the inversion of a PT-oriented Flip-Flop cassette, inserted into the coding intron of a hypothetical gene. Upon FLP-expression, the FLEx switch undergoes two recombination events: (1) recombination between the two FRT sites or between the two FRT14 sites leads to cassette inversion that is followed by (2) excision of either the pair of FRT sites or the pair of FRT14 sites, both of which have obtained the same orientation due to the flip. Since the remaining unpaired FRT and FRT14 sites are not able to recombine, the cassette will be locked in the GT orientation. Thus, the initial PT orientation allows the gene to be tracked by EGFP-tagged protein expression in tissues. FLP activity inverts the Flip-Flop cassette in random cells, generating a mosaic tissue consisting of cells that did not undergo the flip and are still expressing the EGFP-tagged protein and cells that inverted the Flip-Flop cassette into the GT orientation, which are marked by mCherry expression. (C) (a) Wing imaginal disc from a eff MI05507EGFP-T2A-mCherry /+ third instar larva, expressing Eff-EGFP-Eff (b) Wing imaginal disc from a hsFLP; eff MI05507EGFP-T2A-mCherry /+ third instar larva showing hsFLP-induced clones that loose the GFP signal (c) and express mCherry ( . Hence, the PT orientation insertions permit gene expression analysis and can effectively be inverted to create mCherry marked mutant cells. A comparison of the PTs generated using a previous, shorter, RMCE construct (GFSTF; Venken et al., 2011) and the PT of the Flip-Flop cassette did not show any obvious difference in expression pattern or genetic properties based on complementation tests. In summary, if the protein traps are functional with the GFSTF cassette they are also functional with the EGFP-tagged proteins derived from the Flip-Flop.
In addition to the above-described PT lines, we generated GT-lines expressing mCherry for five genes: eff, Cdep, Trim9, Ank2, and Ctrip. Complementation tests with independently derived lossof-function alleles (or deficiencies) indicate that these GT-alleles are indeed loss-of-function alleles. Note that none of these GT insertions induce an obvious dominant phenotype. We observed efficient conversion of an initial GT insertion to the PT orientation for eff MI05507 T2A-mCherry-EGFP , using either the ey-FLP or hsFLP drivers in larval brains. As shown, the use of ey-FLP introduces Eff-EGFP-Eff expression in the eye discs (area within white dotted lines), as well as the area of the brain where eyeless is expressed (demarcated by blue dotted lines) (Figure 1-figure supplement 4A). In contrast to the ey-FLP, smaller clones of randomly positioned wild-type and mutant cells can be generated by stochastic FLP expression when applying heat shocks. Indeed, heat-shock-induced hsFLP expression brought about small patches of EGFP-positive cells in eff MI05507 T2A-mCherry-EGFP larval brains ( Figure 1-figure supplement 4B). This type of experiment creates wild-type cells in an otherwise mutant animal and should help to assess whether a mutant phenotype can be attenuated or reverted, or not. In addition, these experiments may help identify in which tissue an essential gene is required.
Finally, for Nedd8 and Sik3 we generated PT lines, but these insertions failed to complement corresponding deficiencies, showing that the internal EGFP tag disrupts protein function. Nevertheless, we were able to induce inversions for both genes using ey-FLP, confirming our observation that flipping the Flip-Flop cassette is efficient (Figure 1 To examine and compare the efficiency of the Flip-Flop inversions, we tested six genes using the same FLP-source, under identical conditions. For SNF4Ag, Trim9 and eff, more than 95% of the cells undergo cassette inversion. For Nedd8, cdep, and Sik3, more than 70% cells displayed Flip-Flop inversions (Figure 1-figure supplement 5). Hence, inducing Flip-Flop inversions in eye-antennal discs is highly efficient for all genes tested.

Conditional gene inactivation using Flip-Flop in post-mitotic cells
Next, we tested if Flip-Flop can be used for gene inactivation in post-mitotic cells by using Trim9 as an example. Previously, Song et al. (2011) created an imprecise excision of a P-element (P{GawB} Trim9[NP4638]) inserted in the 5 0 -regulatory region of Trim9 and reported that animals lacking the first few amino acids of Trim9 are viable and exhibit a droopy wing phenotype (Song et al., 2011). We generated both GT-(Trim9 MI12525P T2A-mCherry-EGFP ) and PT-oriented (Trim9 MI12525EGFP-T2A-mCherry ) Flip-Flop insertions in Trim9 (Figure 2-figure supplement 1A a). The GT insertion is homozygous lethal and does not complement a deficiency (Trim9 Df ), while the PT insertion is viable in trans over Trim9 Df (Figure 2-figure supplement 1A b) suggesting that the PT encodes a functional protein. When the PT insertion in Trim9 MI12525EGFP-T2A-mCherry /Trim9 Df animals was inverted in all cells using a ubiquitously expressed (Actin-GAL) FLP, the animals died as second instar larvae. This, together with the lethality observed in the GT insertion, suggests that a strong loss of Trim9 leads to early developmental lethality.
We then examined the consequences of neuronal loss of Trim9, by inverting the Trim9 MI12525EGFP T2A-mCherry PT insertion in larval and adult brains using neuronally expressed FLP (nSyb-GAL4>UAS-FLP) (Figure 1-figure supplement 2C and Figure 2-figure supplement 1B). Prior to the flip, in the PT orientation, Trim9 is expressed in numerous neurons in the larval brain (Figure 2A a-d) and the protein is localised to nuclei and cytoplasm (Figure 2A e-h). These cells do not express mCherry. Following inversion to the GT orientation, we detected mCherry expressing cells in the larval brain (Figure 2A a'-d'). Upon close examination, we observed three different populations of neurons. First, cells that have undergone cassette inversion and are marked with mCherry ( Figure 2A (3) Note that Trim9-EGFP-Trim9 levels are lower in these cells. We reason that these cells have undergone cassette inversion and are hence expressing mCherry, but that the Trim9-EGFP-Trim9 that was produced prior to the flip has not been fully degraded. Thus, the population of cells that express only mCherry must have lost most of the tagged Trim9 protein, whereas Trim9-EGFP-Trim9 perdures in a subset of cells that are yellow. Importantly, this allows analysis of protein perdurance. In addition, it permits researchers to only select and analyse those cells that are truly mutant and have lost the majority of the wildtype protein. Neuronal Trim9 inactivation does not affect the eclosion of the animal but leads to the sterility of short-lived animals with droopy wings, corroborating previous observations associated with the loss of Trim9 (Song et al., 2011).
Next, we explored the use of the Flip-Flop cassette in post-mitotic, adult neurons. We expressed GAL80 ts (a suppressor of GAL4 at low temperature) to prevent nSyb-GAL4 from inducing flips during development. tubP-GAL80 ts ; Trim9 MI12525EGFP-T2A-mCherry/ Trim9 Df ; UAS-FLP/nSyb-GAL4 animals were raised at 18˚C until the fourth day of adulthood and were subsequently shifted to 29˚C. Animals that were not shifted to 29˚C did not express mCherry in adult brains ( Figure 2B a-d). In contrast, those shifted to a higher temperature on the fourth day of adulthood showed loss of Trim9-EGFP-Trim9 expression and gain of mCherry in a large subset of neurons ( Figure 2B a'-d'). Interestingly, animals shifted to 29˚C do not show droopy wing phenotype, but have a reduced lifespan. This suggests that the droopy wing phenotype observed by Song et al. (2011) is due to the loss of neuronal Trim9 during development, and not because of loss of the gene in adults.

Loss of SNF4Ag in adults flies leads to severe neurodegeneration
To further explore the use of Flip-Flop in adult flies, we selected a subunit of the AMP-activated protein kinase (AMPK) complex. The AMPK complex is involved in sensing stresses such as a drop in the ratio of ATP/AMP, hypoxia, ischemia, and heat-shock (Hardie et al., 2003). The ATP/AMP ratio is sensed through an allosteric mechanism by the non-catalytic g subunit of the AMPK complex, which in turn promotes the phosphorylation of the catalytic a subunit enhancing a subunit's kinase activity (Sanders et al., 2007;Scott et al., 2004). Phosphorylation of AMPK's targets by the a-subunit activates a signalling cascade that ultimately regulates fatty acid oxidation, autophagy, and mitochondrial biogenesis and boosts the ATP/AMP ratio within the cell (Hardie et al., 2012). Sensing the ATP/AMP ratio by the g subunit is thus critical for the proper function of the AMPK complex, and together with the other AMPK subunits, the g subunit is crucial for proper energy homeostasis (Hardie et al., 2003).
Next, we induced Flip-Flop inversion in eyes. The Drosophila eye is a highly organised structure that is composed of about 800 small functional units called ommatidia. Every ommatidium contains eight photoreceptor cells (R1-R7) each with a rhabdomere, the light-sensing organelle. Within each ommatidium, rhabdomeres are arranged in a highly stereotypic fashion ( Figure 3E c control panel, indicated by numbers). In ey-FLP;;SNF4Ag MI09417EGFP-T2A-mCherry /SNF4Ag Df animals, almost all photoreceptors inverted from the PT-to GT-orientation, as revealed by mCherry expression (Figure 3E b'). Inverting the cassette in SNF4Ag to a GT orientation reduces the amplitude of the electroretinogram (ERG) in 1-day-old flies. These animals did not show a stronger reduction in ERG amplitude when measured at day 30 and hence do not show further degeneration ( Figure 3C and D). We also observed a reduction in the number of rhabdomeres as well as defects in photoreceptor organisation ( Figure 3E-c'). The loss of SNF4Ag disrupts energy homeostasis, possibly prevents the photoreceptors from meeting their high-energy demand and leading to a reduction in their activity. Similar defects have been described in alicorn mutant eye clones (SNF4Aa) (Spasic´et al., 2008).
It has been previously shown that SNF4Ag is required for autophagy (Lippai et al., 2008). We, therefore, analysed the role of SNF4Ag in autophagy in mutant cells generated by Flip-Flop inversion. To this end, we induced Flip-Flop-mediated loss of SNF4Ag in the adult gut and starved adults for 24 hr by growing them on agar lacking amino acids (aa) ( Figure 4A). To analyse the induction of autophagy, we stained adult gut from these animals for p62, a marker that is degraded by autophagy. If SNF4Ag is required for autophagy, loss of SNF4Ag should lead to p62 accumulation in mutant cells, as they should not induce autophagy. However, we observed a loss of p62 in mutant cells (mCherry positive) and robust staining of p62 in control cells (EGFP positive, indicated by red arrows) ( Figure 4B and Figure 4-figure supplement 1). Interestingly, this parallels the observation by Sahani et al. (2014) in vertebrate cells. These authors observed that following prolonged amino acid starvation, the autophagy marker p62 is restored because of an increase in transcription and translation mediated by the availability of autophagy derived aa. Hence, we propose that loss of SNF4Ag blocks autophagy, deprives cells of aa during starvation, leading to a loss of p62. Similar data have been documented in Drosophila larvae and mammalian cells (B'chir et al., 2014;Colosetti et al., 2009;Duran et al., 2011;Erdi et al., 2012).
Previous studies have described that developmental loss of SNF4Ag leads to neurodegeneration (Cook et al., 2012;Tschäpe et al., 2002). To further characterise the role of SNF4Ag in Drosophila neurons, we used a post-mitotic GAL4 driver C155-GAL4, to drive UAS-FLP and obtain GT orientated Flip Flops in all neurons. Such pan-neuronal flip of SNF4Ag MI09417EGFP-T2A-mCherry into a nonfunctional GT leads to severe locomotor defects in adult flies (data not shown). We also examined the role of SNF4Ag in adults using hsFLP-mediated cassette inversion. Flies of the genotype hsFLP; SNF4Ag MI09417EGFP-T2A-mCherry / SNF4Ag MI09417 were raised at 18˚C and were given a 2-hr heatshock at 37˚C, 3 days after eclosion. Two days after the heat-shock, H and E staining of adult brains revealed severe neurodegeneration. The extent of neuronal loss did not worsen by day 10 ( Figure 4C b and d). In addition, these flies typically die 3 to 4 weeks later. Following heat-shock, animals that do not carry hsFLP do not show obvious lesions in the brain ( Figure 4C a and c). We conclude that loss of SNF4Ag leads to neuronal demise in adult flies.
In conclusion, the Flip-Flop strategy allows conditional gene inactivation and generation of mosaics for a side-by-side comparison of mutant and wild-type cells in the same tissue, while simultaneously marking wild-type and mutant cells with different fluorescent markers. By using different sources of FLP, one can control the spatial and temporal pattern of mosaic generation and gene inactivation. Using hsFLP, we show that Flip-Flop can be used to severely inhibit gene function during development, as well as in adults in both mitotically active and post-mitotic cells. Flip-Flop can therefore reliably be used to explore gene functions both during development and in adult flies. Recently, a similar Flippase-based conditional gene inactivation method 'FlpStop' was published (Fisher et al., 2017). Both the FlpStop and Flip-Flop strategies employ a similar Flex switch to induce local cassette inversion. However, there are several important differences between the two technologies. (1) FlpStop does not tag the endogenous protein with a fluorescent tag, while Flip-Flop produces an internally EGFP-tagged protein. While this may seem beneficial for genes where an internal tag is deleterious to the protein, we argue that for the majority of the genes the internal EGFP tag is not detrimental (this study; Nagarkar-Jaiswal et al., 2015). In contrast, endogenous tagging is very advantageous as it permits to track the expression and subcellular localization of a protein, allows anti-GFP antibody-mediated immunoprecipitations followed by mass spectroscopy (David-Morrison et al., 2016;Yoon et al., 2017), enables ChIP sequencing (Nègre et al., 2011), and permits mRNA or protein removal with iGFPi (Neumüller et al., 2012) or deGradFP (Caussinus et al., 2011;Nagarkar-Jaiswal et al., 2015;Urban et al., 2014). Finally, the EGFP tag allows us for the first time to track protein perdurance in mutant cells.
(2) The second major difference between FlpStop and Flip-Flop relates to how mutant cells are marked. In FlpStop, tdTomato is expressed under the control of the UAS/GAL4 system. Hence, cells that have undergone inversion will express tdTomato, irrespective of the expression pattern of the gene. In contrast, Flip-Flop will label cells with mCherry only if they express the gene of interest. In summary, the ability to track the loss of the protein of interest in mutant cells concomitant with loss of the EGFP signal while simultaneously marking the mutant cells with mCherry from the same regulatory elements distinguishes Flip-Flop from all other mosaic analyses methods.