MAGIC relies on a gRNA-marker dual-transgene (Figure 1A, Figure 1—figure supplement 1B) inserted in a specific chromosomal arm to both induce and visualize clones homozygous for this arm (Allen et al., 2021). The gRNA part of the construct expresses two gRNAs ubiquitously to target a pericentromeric location on this arm. Two gRNAs targeting two sequences that are close to each other, instead of a single gRNA, are used to increase the probability of DSBs and thus the clone frequency. The marker part in pMAGIC utilizes a ubiquitously expressed Gal80 to prevent Gal4-dependent labeling of heterozygous and homozygous cells for the gRNA-marker, while allowing labeling of homozygous cells that lose the gRNA-marker. In contrast, nMAGIC uses ubiquitously expressed BFP, which results in brighter labeling of gRNA-marker homozygous cells, intermediate labeling of gRNA-marker heterozygous cells, and lack of labeling of gRNA-marker-negative homozygous cells (Figure 1—figure supplement 1B).

We have previously developed gRNA-marker vectors for both pMAGIC and nMAGIC. However, gRNA-markers made with these vectors exhibit some limitations. First, the clone frequency can be low for certain gRNAs (Allen et al., 2021). Second, the ubi-Gal80 in pMAGIC gRNA-markers does not completely suppress Gal4 activity in certain tissues, such as hemocytes. Third, pMAGIC clones are sometimes too dim to visualize cell morphology, such as thin dendrite and axon projections of neurons. Lastly, nMAGIC utilizes a nuclear BFP marker (nlsBFP), which does not show the cell shape and sometimes cannot mark clones effectively. Thus, to optimize MAGIC, we first sought to improve the gRNA-marker designs.

Since clone induction in MAGIC depends on gRNA-induced DNA DSBs, we asked whether a more efficient gRNA design enhances clone frequency in somatic tissues. The previous gRNA-marker vectors used a tgFE design, which contains a flip of A-U positions and a stem-loop extension (F+E) of the original gRNA scaffold and a tRNA^Gly before each gRNA targeting sequence (Figure 1A; Allen et al., 2021). However, we have shown that an improved Qtg2.1 design, which contains an additional extension of the second stem loop in the gRNA scaffold (gRNA2.1) and a single tRNA^Gln spacer between the two gRNAs (Figure 1A), is much more mutagenic than tgFE in somatic tissues (Koreman et al., 2021). We thus compared two pMAGIC gRNA markers that are based on these two gRNA designs but target the same genomic sequences at cytological band 40D2 to assess their ability to induce clones in peripheral sensory neurons. When used with the same neuronal/epidermal precursor Cas9, zk-Cas9 (Allen et al., 2021), Qtg2.1 resulted in three times more neuronal clones as compared to tgFE (Figure 1B), confirming a positive correlation between gRNA efficiency and clone frequency.

 To ensure more complete suppression of Gal4 activity by Gal80 in pMAGIC, we replaced the ubi enhancer driving Gal80 expression with an αTub84B (tub) enhancer that was used in tub-Gal80 in the MARCM system (Lee and Luo, 1999). When combined with the larval hemocyte marker pxn-Gal4 UAS-CD4-tdTom (Han et al., 2014; Figure 1C), ubi-Gal80 did not completely suppress the labeling of hemocytes (Figure 1D). In contrast, no labeled hemocytes could be detected in the presence of tub-Gal80 (Figure 1E), suggesting that tub-Gal80 is a better marker for pMAGIC.

To improve clone brightness in pMAGIC, we sought to destabilize Gal80 and reduce its expression, reasoning that dim clones are due to prolonged Gal80 activity after clone induction. In addition to replacing the ubi enhancer with the tub enhancer, we also introduced protein and mRNA destabilization sequences (DE) (Li et al., 1998; Zubiaga et al., 1995) at the 3’ end of the Gal80 coding sequence, and replaced the His2Av polyA by SV40 polyA (Figure 1F), as the latter reduces transgene expression (Han et al., 2011). By measuring the brightness of epidermal (Figure 1G) and neuronal (Figure 1H) clones induced by gRNA-40D2 and gRNA-42A4, we found that each of the three changes improved clone brightness.

Finally, to visualize cell shape in nMAGIC, we replaced nlsBFP with cytosolic BFP tagged by 3 X Hemagglutinin (HA), in addition to utilizing the tub enhancer. This new design allowed us to better discern clone shapes in both the wing imaginal disc and the epidermis (Figure 1J, J’ , and L), in contrast to the previous ubi-nlsBFP design (Figure 1I and K). To make nMAGIC compatible with more fluorescent reagents, we generated an additional vector that contains miRFP680, a far-red/infrared fluorescent protein (IFP) (Matlashov et al., 2020), in place of BFP. Wing-disc clones labeled with this marker were readily detectable in unstained tissues (Figure 1M). By measuring the sizes of homozygous gRNA-marker clones (BFP/BFP) and homozygous wild-type (WT) clones (+/+) in wing discs, we found that these two cell populations in the twin spots showed no noticeable bias in growth or viability (Figure 1O).

Thus, by altering the designs of the gRNAs and the Gal80 and BFP markers, we created new vectors optimized for more robust applications of nMAGIC and pMAGIC.

To enhance the utility of MAGIC in Drosophila, we generated complete sets of pMAGIC and nMAGIC (BFP version) gRNA-markers for all four chromosomes (Figure 2A). To identify suitable gRNA target sites, we analyzed the pericentromeric sequences of X, 2 L, 2 R, 3 L, 3 R, and 4, based on three criteria (Allen et al., 2021): (1) conserved in closely related Drosophila species to minimize the chance of single nucleotide polymorphism, (2) located away from functionally critical regions to avoid disrupting essential processes, and (3) unique within the genome to minimize off-target effects. For each MAGIC construct, we selected a pair of non-repetitive gRNA target sequences in intergenic regions to maximize the chances of DSBs. These two sequences are closely linked to minimize the risk of large deletions. Given the variable efficiency of gRNA target sequences, we selected three pairs of gRNAs targeting three chromosomal locations for each chromosomal arm and named them according to the corresponding cytoband (Table 1).

To evaluate the clone-induction properties of these gRNAs, we combined the pMAGIC set with zk-Cas9 and counted the number of neuronal clones in A1-A7 larval hemi-segments (Figure 2B). As expected, the clone frequency varied from gRNA to gRNA, but we were able to identify efficient gRNAs (≥ 10 clones per larva) for every chromosomal arm. We previously found that different gRNAs follow the same trend of relative efficiency in different tissues (Allen et al., 2021). Here we additionally tested nMAGIC gRNAs for 2 R in the wing disc (Figure 2C) and noticed a similar trend of clone induction to that of their pMAGIC counterparts in sensory neurons (Figure 2B), suggesting that the results in sensory neurons are transferrable to other tissues.

Certain gRNAs (e.g. gRNA-40E1) exhibited very low clone frequency. Such gRNAs can be useful for inducing sparse clones in highly packed tissues such as the brain. For example, using the same ey-Cas9 (Ji et al., 2022), the highly efficient gRNA-40D2 induced too many neuronal clones in the adult brain for morphological analysis (Figure 2D), while gRNA-40E1 gave rise to few clones, whose projection patterns were much easier to analyze (Figure 2E).

Together, the pMAGIC and nMAGIC gRNA-marker lines constitute a complete kit (Table 1) for genome-wide MAGIC applications in Drosophila.

To determine if MAGIC can be applied to diverse tissues in Drosophila, we conducted clonal analysis in the larva using gRNA-40D2(Gal80) and several tissue-specific Cas9s. With the ubiquitous vas-cas9 (López Del Amo et al., 2022) and tub-Gal4 >UAS-mCD8-GFP, we readily detected clones in the larval brain (Figure 3A), proliferating tissues like eye and leg discs (Figure 3B–C), and polyploid tissues like the fat body, gut, and trachea (Figure 3D–F). Clone induction in polyploid tissues suggests that crossing-over events occurred before the last cell division. Using zk-Cas9 and R38F11-Gal4>UAS-tdTomato (tdTom), we observed frequent epidermal clones (Figure 3G). Using the glia precursor gcm-Cas9 and repo-Gal4 >UAS-mCD8-GFP, we detected individual glial clones in the brain (Figure 3H). The new pMAGIC gRNA-marker design allowed us to reliably induce hemocyte clones (Figure 3I).

The ability to control the timing of clone induction has been instrumental for neuronal birth dating and modulation of clone frequency in traditional MARCM analysis of the Drosophila adult brain. To explore the potential of pMAGIC to serve similar purposes, we used heat-shock (hs) Cas9 (Garcia-Marques et al., 2020), along with gRNA-40D(Gal80) and a pan-neuronal Gal4/UAS-membrane marker combination (RabX4-Gal4>UAS-MApHS), to induce neuronal clones in the adult brain. Heat shock at 120 hr after egg laying (AEL) induced too many clones for separating individual neurons (Figure 3J–J’), while later heat shock at 48 hr after pupal formation (APF) produced many fewer, spatially separated clones in the central brain. These results collectively demonstrate MAGIC’s efficacy and flexibility in generating clones in diverse Drosophila tissues, indicating its value in studying gene function and cell lineage in various tissue contexts.

The neuromuscular junction (NMJ) in Drosophila larvae has been a valuable model for elucidating synaptic biology, owing to its simplicity, accessibility, and conserved features shared with mammalian excitatory synapses (Charng et al., 2014; Menon et al., 2013; Chou et al., 2020). Despite its popularity, the NMJ has rarely been analyzed by the MARCM technique in the literature, likely due to the difficulty of inducing clones in motor neurons. To test the effectiveness of MAGIC in analyzing gene function at the NMJ, we selected two genes crucial for synaptic function and construction: Vesicular glutamate transporter (VGlut) (Daniels et al., 2004) and bruchpilot (brp) (Kittel et al., 2006; Wagh et al., 2006), null mutations of each of which exhibit recessive lethality in larvae. Combined with appropriate gRNA(Gal80) lines (42A4 for VGlut on 2 R and 40D2 for brp on 2 L) and tub-Gal4 UAS-CD8-GFP, zk-Cas9 induced frequent clones in type Ib boutons (Figure 4A–B”). The loss of VGlut or Brp specifically in GFP-labeled NMJs was confirmed by immunostaining. These results exemplify the value of pMAGIC for dissecting gene function in NMJ biology at the single-cell level.

Because all FRT sites in the existing FRT/Flp mosaic system are located at some distances away from centromeres, it has not been possible to study genes located between the FRT sites and the corresponding centromeres by clonal analysis. In contrast, the gRNA target site in a MAGIC experiment can be user-selected to enable clonal analysis of any given gene. To illustrate this potential, we used gRNA-41F9 to induce homozygous mutant clones of Ecdysone receptor (EcR), which is located at 42A10 and is inaccessible to all FRT sites on 2 R. EcR is a transcription factor required for neuronal remodeling during metamorphosis (Brown et al., 2006). In peripheral sensory neurons, EcR activation at the beginning of pupariation causes apoptosis of some dendritic arborization (da) neurons while triggering dendrite pruning of other da neurons (Williams and Truman, 2005). As expected, a wild-type (WT) pMAGIC clone of class IV da (C4da) neuron exhibited complete dendrite pruning at 16 hr after puparium formation (APF), accompanied by dendrite debris phagocytosed into epidermal cells (Han et al., 2014; Figure 5A). In contrast, EcR mutant da neuron clones still maintained larval dendritic arbors at 16 hr APF (Figure 5B–D), instead of dying (in the case of C3da) or undergoing dendrite pruning (in the cases of C1da and C4da).

Mosaic analysis of genes located on the 4th chromosome had not been possible until the recent introduction of FRT sites onto this chromosome by CRISPR-mediated knock-in (Goldsmith et al., 2022). Despite these advances, existing mutations on FRT-lacking 4th chromosomes still cannot be analyzed by the FRT/Flp system, given that meiotic recombination is exceedingly rare on the 4th chromosome, preventing introduction of FRT sites onto mutant chromosomes. A valuable gene-disruption resource in Drosophila is the deficiency library consisting of strains harboring chromosomal deletions. MAGIC can potentially be used in conjunction with the deficiency library to screen for genes important in a particular biological process. To test the potential of MAGIC for analyzing mutations on the 4th chromosome and for gene discovery with deficiencies, we generated pMAGIC C4da clones of Df(4)ED6380, a deficiency that deletes a segment between cytological bands 102B7 and 102D5. These clones show dramatically reduced dendrites compared to the WT controls (Figure 5E–G), indicating the existence of genes important for dendrite growth in this region.

When examining nMAGIC gRNA-markers for the 4th chromosome, we noticed uneven expression of tub-3xHA-BFP in epidermal cells and some imaginal tissues, exemplified by cells lacking detectable BFP signal (Figure 5—figure supplement 1A and B). This variegated expression is likely due to transgenes residing in heterochromatin and is common for transgenes located on the 4th chromosome (Riddle et al., 2011). While this uneven expression limits the usefulness of our gRNA-markers in the corresponding epithelial tissues, Gal80 in pMAGIC gRNA-markers efficiently suppressed Gal4 activities in most neurons (Figure 5—figure supplement 1C), confirming their effectiveness in neuronal MAGIC analysis.

Lastly, we wondered if the MAGIC system can generate clones in interspecific hybrid animals derived from D. melanogaster and D. simulans parents, given that the two species show a large degree of synteny (Chakraborty et al., 2021). The clones in these animals would contain homozygous chromosomal arms derived from a single species. To test this idea, we crossed D. melanogaster females containing gRNA-42A4(BFP); hh-cas9 to D. simulans males carrying a loss of function mutation of Lethal hybrid rescue mutation (Lhr), which ensures the viability of the hybrid male progeny (Barbash, 2010). Indeed, we observed twin spots consisting of dark clones, containing only D. simulans 2 R, and brighter clones, containing only D. melanogaster 2 R, in wing discs of both female and male progeny (Figure 5G and H), demonstrating the feasibility of studying species-specific alleles in cell-cell interactions in interspecific hybrids.

See the Key Resource Table for details of fly stocks used in this study. Most fly lines were either generated in the Han lab or obtained from the Bloomington Drosophila Stock Center. lethal hybrid rescue (lhr) mutant D. simulans was a gift from Dr. Dan Barbash. All flies were grown on standard yeast-glucose medium, in a 12:12 light/dark cycle, at 25 °C unless otherwise noted. Virgin males and females for mating experiments were aged for 3–5 days.

To generate and label pMAGIC clones in larval peripheral sensory neurons, we used either RabX4-Gal4 UAS-MApHS (for gRNA-markers on chromosomes X, II, and IV) or 21–7 Gal4 UAS-MApHS (for gRNA-markers on chromosome III) combined with zk-cas9. To count peripheral sensory neuronal clones on the larval body wall, third instar larvae were mounted laterally on slides and then counted in segment A1-A7 under a Nikon SMZ18 stereomicroscope. pMAGIC clones were induced and labeled by RabX4-Gal4 UAS-MApHS combined with ey-cas9 or hs-cas9 in the fly adult brain, by tub-Gal4 UAS-mCD8-GFP combined with vas-cas9 in larval brains, imaginal discs, fat bodies, guts, and trachea, by repo-Gal4 UAS-mCD8-GFP combined with gcm-cas9 in glia, by pxn-Gal4 UAS-tdTom combined with Act-cas9 in hemocytes, by tub-Gal4 UAS-mCD8-GFP combined with zk-cas9 in larval motor neurons, by R38F11-Gal4 UAS-tdTom combined with zk-cas9 in the larval epidermis. To induce nMAGIC clones in wing imaginal discs of interspecific hybrid animals, we use gRNA-42A4(BFP); hh-cas9 virgin females of D. melanogaster to cross with Lhr¹ (Brideau et al., 2006) D. simulans males.

Live imaging of larval epidermal cells, sensory neurons, and hemocytes was performed as previously described (Poe et al., 2017). Animals were collected at 96 (for late third larvae) or 120 hr (for wandering third instar larvae) AEL and mounted in glycerol on a slide with vacuum grease as a spacer. Animals were imaged using a Leica SP8 confocal microscope with a 40 X NA1.3 oil objective, pinhole size 2 airy units, and a z-step size of 1 µm. For the epidermis, images were taken at the dorsal midline of A2 and A3 segments. For dendritic arborization neurons, images were taken from A1 to A7 hemi-segments.

For imaging sensory neurons in pupae, newly formed pupae were collected and incubated at 25 °C. After 16 hr, the pupal cases were carefully removed, and the pupae were mounted dorsal side up on slides with halocarbon oil beneath a coverslip. Double-sided tape was used as a spacer. The mounted pupae were imaged using a Leica SP8 confocal microscope with a 40 X NA1.3 oil objective.

To induce heat shock, vials containing animals at the appropriate developmental stages were submerged in a 37 °C water bath for 1 hr. Subsequently, the vials were transferred back to a 25 °C incubator until adults emerged.

Larval dissections were performed as described previously (Poe et al., 2019). Briefly, wandering third instar larvae were dissected in a small petri dish filled with cold phosphate-buffered saline (PBS). The anterior half of the larva was inverted. To prepare imaginal discs, trachea, and gut were removed. Samples were then transferred to 4% formaldehyde in PBS and fixed for 20 min at room temperature. After washing with PBS, the tissues were stained in DAPI (1:1000) in 0.2% PBST (PBS with 0.2% Triton X-100) for 5 min. The tissues were washed again in PBST and mounted in SlowFade Diamond Antifade Mountant (Thermo Fisher Scientific) on a glass slide. A coverslip was lightly pressed on top. Imaginal discs were imaged using a Leica SP8 confocal microscope with a 20 X NA0.8 oil objective.

Flies were aged for 1 day after eclosion. Brains were dissected in PBS at room temperature and then fixed in 4% paraformaldehyde in PBS with constant circular rotation for 20 min at room temperature. The brains were subsequently washed in 0.2% PBST and mounted on a glass slide under a glass coverslip. Vacuum grease was used as a spacer between the coverslip and the slide. Brains were imaged using a Leica SP8 confocal microscope with a 40 X NA1.3 oil objective.

Larval fillet dissection was performed on a petri dish half-filled with PMDS gel. Wandering third instar larvae were pinned on the dish in PBS dorsal-side up and then dissected to expand the body wall. PBS was then removed, and 4% formaldehyde in PBS was added to fix larvae for 15 min at room temperature. For VGlut staining, the fillets were fixed in Bouin’s solution for 5 min at room temperature. Fillets were rinsed and then washed at room temperature in PBS for 20 min or until the yellow color from Bouin’s solution faded. After immunostaining, the head and tail of fillets were removed, and the remaining fillets were placed in SlowFade Diamond Antifade Mountant on a glass slide. A coverslip was lightly pressed on top. Larval fillets were imaged using a Leica SP8 confocal microscope with a 40 X NA1.3 oil objective.

Larval brains and larval fillets were rinsed and washed at room temperature in 0.2% PBST after fixation. The samples were then blocked in PBST with 5% normal donkey serum (NDS) for 1 hr before incubating with appropriate primary antibodies in the blocking solution. Brains were stained for 2 hr at room temperature, and fillets were stained overnight at 4 °C. After additional rinsing and washing, the samples were incubated with secondary antibodies for 2 hr at room temperature. The samples were then rinsed and washed again before mounting and imaging. Primary antibodies used in this study are mouse anti-Repo antibody 8D12 (1:50 dilution), mouse anti-Brp antibody nc82 (1:100 dilution), rabbit anti-VGlut (1:200 dilution; Chen et al., 2024), mouse anti-HA antibody (12CA5, 1:100), and goat anti-HRP conjugated with Cy3 (1:200). Secondary antibodies include donkey anti-mouse antibody conjugated with Cy5 (1:400) and donkey anti-rabbit antibody conjugated with Cy5 (1:400).

ImageJ analyses were conducted in Fiji/ImageJ. To compare brightness of clones in sensory neurons and epidermal cells, clones were detected based on thresholds to generate masks. The clone brightness within the masks was then measured as mean pixel intensity. To quantify clones in wing discs, two to three slices of optical sections in the middle of the disc were projected into a two-dimensional (2D) image. Clones were detected based on a fixed threshold to generate masks, and the total area of clones in the wing pouch of each disc was measured.

The tracing and measurement of the neuron dendrites were done as previously described in detail (Poe et al., 2017). Briefly, dendrites were segmented using local thresholding. The segments were then converted into single-pixel-width skeletons. The total length of skeletons was calculated based on pixel distance. Normalized dendrite length was calculated as dendritic length (μm)/segment width (μm).

One-way analysis of variance (ANOVA) with Tukey’s honest significant difference (HSD) test was used when the dependent variable was normally distributed and there was approximately equal variance across groups. A Student’s t-test or paired t-test was used when two groups were compared. For additional information on the number of samples, see figure legends. R Studio was used for all statistical analyses.