The Classic Lobe Eye Phenotype of Drosophila Is Caused by Transposon Insertion-Induced Misexpression of a Zinc-Finger Transcription Factor

Abstract

Drosophila Lobe (L) alleles were first discovered ∼100 years ago as spontaneous dominant mutants with characteristic developmental eye defects. However, the molecular basis for L dominant eye phenotypes has not been clearly understood. A previous work reported identification of CG10109/PRAS40 as the L gene, but subsequent analyses suggested that PRAS40 may not be related to L. Here, we revisited the L gene to clarify this discrepancy and understand the basis for the dominance of L mutations. Genetic analysis localized the L gene to Oaz, which encodes a homolog of the vertebrate zinc finger protein 423 (Zfp423) family transcriptional regulators. We demonstrate that RNAi knockdown of Oaz almost completely restores all L dominant alleles tested. L^rev6-3, a revertant allele of the L² dominant eye phenotype, has an inframe deletion in the Oaz coding sequence. Molecular analysis of L dominant mutants identified allele-specific insertions of natural transposons (roo[ ]L¹, hopper[ ]L⁵, and roo[ ]L^r) or alterations of a preexisting transposon (L²-specific mutations in roo[ ]Mohr) in the Oaz region. In addition, we generated additional L²-reversion alleles by CRISPR targeting at Oaz. These new loss-of-function Oaz mutations suppress the dominant L eye phenotype. Oaz protein is not expressed in wild-type eye disc but is expressed ectopically in L²/+ mutant eye disc. We induced male recombination between Oaz-GAL4 insertions and the L² mutation through homologous recombination. By using the L²-recombined GAL4 reporters, we show that Oaz-GAL4 is expressed ectopically in L² eye imaginal disc. Taken together, our data suggest that neomorphic L eye phenotypes are likely due to misregulation of Oaz by spontaneous transposon insertions.

IT has long been proposed that natural transposons can affect the genome structure and gene regulation in diverse species from plants to humans (Miller and Capy 2004; Maksakova et al. 2006; Burns and Boeke 2012). Transposons have conferred important sources for spontaneous mutations and genetic polymorphisms, driving host genome evolution. Transposable elements can be classified into two major groups by transposition mechanisms: DNA transposons and retrotransposons (Bourque et al. 2018). In Drosophila, DNA transposons such as P and Hopper elements transpose through a cut-and-paste mechanism. In contrast, retrotransposons like gypsy and opus are mobilized by a copy-and-paste mechanism. Both groups of natural transposons have been identified abundantly throughout different strains of Drosophila melanogaster, composing 10–20% of the whole genome with diverse subgroups (Kaminker et al. 2002; Kapitonov and Jurka 2003; Ganko et al. 2006).

Studies on several spontaneous mutants have shown that natural transposons can generate loss- or gain-of-function mutations in developmentally important genes by insertions into coding sequences or critical regulatory sequences near the affected genes. For example, Glazed (Gla) is a dominant mutation that causes defects in eye development. Gla turns out to be a gain-of-function allele of the wingless (wg) gene, and its eye phenotype results from the Wg misexpression induced by transposon sequences of a roo retrotransposon inserted in the wg promoter (Brunner et al. 1999). Other transposon sequences can also drive ectopic transcription of different genes to affect eye development in Drosophila (Awasaki et al. 1996; Mozer 2001). Alternatively, transposons may affect host transcription by altering the chromatin structure around their insertion sites (Morgan et al. 1999; Ong-Abdullah et al. 2015). Hence, transposable elements provide versatile docking sites for DNA-binding proteins to affect host gene expression (Chuong et al. 2017). In this study, we show that several L dominant alleles are caused by insertion of transposable elements or modifications of an existing transposable element.

Classical L dominant alleles are one of the first Drosophila mutants with developmental eye defects. An interesting feature of these eye phenotypes is the preferential reduction in the ventral eye region with variable expressivity depending on genetic backgrounds (Lindsley and Zimm 1992). Genetic interaction studies have shown that L eye phenotype can be modified by genetic changes in diverse signaling pathways, including Notch (N), Wg, Decapentaplegic (Dpp), JAK/STAT, and apoptotic pathways (Chern and Choi 2002; Singh et al. 2005, 2006; Wang and Huang 2009). However, the molecular nature of the dominant L mutations and the basis of these genetic interactions have not been clearly defined.

Previously, we have reported that L is an essential gene for viability identified as an uncharacterized CG10109 (Chern and Choi 2002). CG10109 turned out to be the Drosophila homolog of mammalian PRAS40, a negative regulator of mTOR signaling (Sancak et al. 2007; Vander Haar et al. 2007). As described in this study, however, subsequent analyses suggested that L might be different from CG10109. Coincidently, an independent study reported an isolation of a CG10109/PRAS40 knockout mutant that is viable with no obvious developmental defects (Pallares-Cartes et al. 2012). These results led us to reinvestigate the identity of the L gene and the molecular lesions of L mutant alleles. Here, we show genetic and molecular evidence that CG42702/Oaz adjacent to CG10109/PRAS40 is responsible for L mutant eye phenotypes. The Oaz gene is a unique Drosophila homolog of vertebrate Zfp423 family transcriptional factors containing multiple zinc finger (ZF) motifs. Most vertebrate species have a pair of related genes, Zfp423 (OAZ/Ebfaz/ZNF423) and Zfp521 (Evi3/EHZF/ZNF521), that are involved in diverse developmental processes (Tsai and Reed 1997; Hentges et al. 2005; Alcaraz et al. 2006; Gupta et al. 2010; Kamiya et al. 2011; Kiviranta et al. 2013). A previous study reported that Drosophila Oaz is expressed in the developing and adult brain (Potter and Luo 2010; Hammonds et al. 2013), but only a developmental function of the gene has been reported in the embryonic posterior spiracle development (Krattinger et al. 2007).

Several classical L mutants isolated many decades ago have a common characteristic of dominant eye-specific phenotype. An intriguing question is how spontaneous L mutations are related to the dominant small eye phenotype. Evidence suggests that dominant L eye phenotypes result from gain-of-function mutations of a gene. For example, the dominant L²/+ eye phenotype was reverted by introduction of loss-of-function mutations like chromosomal deficiencies generated by X-ray irradiation (Baker and Ridge 1980; Davis and MacIntyre 1988). However, the molecular basis underlying the dominant gain-of-function phenotypes of the L mutations has been unknown. As described in this study, we have identified natural transposons inserted in the Oaz region of spontaneous L mutants, implying a possible relationship between the transposon insertions and the L eye phenotypes.

Here, we demonstrate that L dominant eye phenotypes are restored by reducing Oaz expression. We also confirm that L dominant eye phenotypes are suppressed by CRISPR targeting of the Oaz gene. Finally, we provide evidence that Oaz protein and Oaz-GAL4 reporter expression are induced ectopically in the mutant eye disc. Taken together, we identify Oaz as the L gene, and propose that spontaneous transposon insertions or an altered transposon in Oaz may drive ectopic L expression in the eye disc to cause the gain-of-function eye phenotypes.

Materials and Methods

Drosophila strains

L dominant mutations (L¹, L², L⁴, L⁵, L^r, and L^si) and ft¹ were described by Lindsley and Zimm (1992) (Bloomington Drosophila Stock Center, BDSC 318, 319, 320, 321, 323, 324, and 304, respectively). L^rev6-3 and PRAS40^P17 were described by Chern and Choi (2002). L^fee (Wang and Huang 2009) and PRAS40^KO (Pallares-Cartes et al. 2012) were obtained from Min-Lang Huang and Aurelio A. Teleman, respectively. Breakpoints of chromosomal deficiencies used for complementation tests (Figure 1A) are described in FlyBase (Thurmond et al. 2019) and the following literature: Df(2R)ED2354 (Ryder et al. 2007), Df(2R)BSC357, Df(2R)BSC668, Df(2R)BSC700 (Cook et al. 2012), Df(2R)Exel8059, Df(2R)BSC11 (Parks et al. 2004), Df(2R)L48 (Davis and MacIntyre 1988), Df(2R)trix (Belote and Baker 1987; Lindsley and Zimm 1992), Df(2R)03072 (Prout et al. 1997) (BDSC 8913, 24381, 26520, 26552, 7877, 6455, 6139, 1896, and 5422, respectively). The following RNAi strains and mutants were obtained from BDSC, Drosophila Genomics Resource Center (DGRC), Vienna Drosophila Resource Center (VDRC), and Harvard Exelixis Collection (Exelixis): Oaz RNAi (BDSC 25923, 35715, VDRC v107061), Oaz-GAL4 (BDSC 30026, DGRC 104937), Oaz-EP (BDSC 43537, Exelixis d09084), PRAS40 RNAi (VDRC v38338, v38339), phyl RNAi (BDSC 29433, VDRC v35469), phyl²²⁴⁵ (BDSC 5686, DGRC 108363), phyl²³⁶⁶ (BDSC 30723), and Achl RNAi (BDSC 25956) (Figure 1B, Figure 2, and Figure 3B).

Figure 1.

Genomic map around L/Oaz gene and localization of L^rev6-3 lethality. (A) Deficiency chromosomes used in complementation tests. The L^rev6-3 lethality was mapped within the genomic region of 51A2-4 from the distal breakpoint of Df(2R)BSC700 to the proximal breakpoint of Df(2R)Exel8059. A single asterisk (*) indicates that breakpoints were confirmed by sequencing analyses, or predicted by the insertion sites of progenitor P-elements (Parks et al. 2004; Ryder et al. 2007; Cook et al. 2012). The double asterisk (**) indicates the breakpoints determined from polytene chromosome analyses and/or genetic mapping. (B) In the genomic region of 51A2-4, four candidate genes of predicted ORFs were identified; Achl, phyl, L/Oaz, and PRAS40 (adapted from the FlyBase Genome Browser). Deficiencies of molecularly defined breakpoints are shown by thick red lines (Df(2R)ED2354, Df(2R)BSC357, Df(2R)BSC668, Df(2R)BSC700, and Df(2R)Exel8059). Gray triangles indicate P-element insertions in candidate genes. Red bars are small deletions. P[PZ]PRAS40^P17 is the P17 insertion (Chern and Choi 2002).

Figure 2.

Suppression of dominant L eyes by Oaz knockdown. (A–L) Suppression of L²/+ eye phenotype by reducing Oaz levels by ey-GAL4. Control knockdown by Achl RNAi (Ri in short) (B), phyl RNAi (C), PRAS40 RNAi (D), and Oaz RNAi (E and F) did not affect eye development. (G) Control L²ey-GAL4/+ shows a typical L²/+ eye phenotype. Achl RNAi (H), phyl RNAi (I), and PRAS40 RNAi (J) failed to modify the L²/+ phenotype. In contrast, two independent Oaz RNAi lines suppressed the L²/+ phenotype completely (K) or strongly (L). (M–R) Suppression of other L dominant alleles by Oaz RNAi. Heterozygous L¹ and L^si show mild eye phenotypes like anterior dents (M and Q, respectively). L⁵/+ eyes are affected more strongly (O). These dominant eye phenotypes of L¹, L⁵, and L^si were fully suppressed to the wild-type level by Oaz RNAi (N, P, and R, respectively). All eye images were taken from female flies using the ZEISS Axiocam Microscope system. Sex chromosomal genotypes are y¹ v¹/ w¹¹¹⁸ or y¹sc* v¹/ w¹¹¹⁸, except (A, D, and J) that have w¹¹¹⁸, in which L phenotypes tend to be weakly enhanced. Achl Ri (B and H) is P[TRiP.JF01976]. phyl Ri (C and I) is P[TRiP.JF03369]. Refer to Materials and Methods for PRAS40 Ri (D and J). Oaz Ri (E, K, N, P, and R) is P[TRiP.GLV21080]. Oaz Ri(b) (F and L) is P[TRiP.JF01943]. Variations of L eye phenotypes were minimized by backcrosses. Penetrance of the L dominant eye phenotypes was: >90% for L²/+ (G), >70% for L¹/+ (M), 50–70% for L⁵/+ (O), and 60–80% for L^si/+ (Q). For each genotype, >1000 flies were examined in different batches. Suppression of L eye phenotypes by Oaz RNAi (K, N, P, and R) was nearly 100% (n > 500).

Figure 3.

Identification of L^rev6-3 mutation and polymorphisms in L mutants. (A) Localization of the L^rev6-3 deletion in Oaz coding sequence. Oaz peptide sequences (coding product of Oaz-RB) of control stocks and L mutants were deduced from genomic sequencing results. Blue letters represent the multiple ZF motifs found in Oaz sequences. Compared to the FlyBase reference sequences (REF in gray), all the examined stocks including w¹¹¹⁸ showed specific polymorphisms (red letters) in peptide sequences as well as numerous nucleotide polymorphisms. L^rev6-3 has a small deletion of six amino-acids in the 10th ZF motif (blue box), but not in L² or other Mohr strains (ft¹, L¹, L², L⁵, and L^r). This deletion is located within the putative Mad-binding motif (ZF7 to ZF13 in bold blue). (B) Summary of amino-acid polymorphisms identified in deduced Oaz sequences of examined stocks of the control w¹¹¹⁸, ft¹, and L mutants, as compared to the FlyBase reference. Polymorphic amino acid changes are indicated as red characters in (A).

To minimize genetic background effects on L eye phenotypes, L alleles and other strains used for genetic tests were backcrossed for 5–10 generations, except for TRiP RNAi strains, which were generated in the isogenic y¹ v¹ or y¹sc* v¹ genetic backgrounds and showed consistent effects on L dominant alleles. Isogenic control strains of w¹¹¹⁸ background (Ryder et al. 2004) were available from BDSC (5905, 5906, 5907, 5908). The Cas9 sources (Kondo and Ueda 2013; Port et al. 2015) were obtained from BDSC (51323, 54590) and the Fly stock center of the National Institute of Genetics (CAS-0004). The reporter stock of the G-TRACE system (Evans et al. 2009) was from BDSC (28280).

Sequencing analyses

Genomic sequences of the region 51A2-4 from Achl to PRAS40 (Figure 1B) were analyzed for L alleles (L¹, L², L⁴, L⁵, L^r, L^si, L^fee, and L^rev6-3) and control strains (CS, w¹¹¹⁸, and ft¹). In the case of L^rev6-3, genomic sequences were amplified from heterozygotes between the mutant and the control isogenic second chromosome. Multiple clones were sequenced independently and compared with the corresponding sequences from the progenitor L² and the control chromosomes. Transposons in specific L alleles were identified by inverse PCR methods. New lethal L alleles were generated by CRISPR targeting as described below and sequenced in the same manner as L^rev6-3 sequencing.

Generation of PRAS40 RNAi strains

The 501 bp unique PRAS40 sequence of exon 2 to exon 3 was amplified by PCR, using two primers, 5′-gtgtctcgatggccaggaagatgtc-3′ and 5′-cagctctctcggccagcgtgg-3′. The amplified fragment was cloned into pWIZ (Lee and Carthew 2003) to generate UAS-PRAS40 RNAi. Transgenic flies were generated by a standard Drosophila germline transformation method (Bachmann and Knust 2008). Transformants were crossed with the ubiquitous Act-GAL4 driver, and the lines showing no phenotype were selected. The selected lines were able to suppress the phenotype of PRAS40 overexpression, confirming their RNAi effects.

CRISPR targeting

All guide RNA (gRNA) sequences used are listed in Supplemental Material, Table S1, and the targeting sites in the L region are shown by green, red, and orange arrowheads in Figure 5, A and M and Figure S8A.

Figure 5.

Reversion of dominant L eye phenotypes by loss-of-function mutations of L. (A) CRISPR genomic target sites for opus[ ]Mohr, L-RB exon1, and L coding sequence are indicated by green, red, and orange arrowheads, respectively. (B–D) Heterozygous eye phenotypes of classical L mutations of Mohr strains, L¹ (B), L⁵ (C), and L² (D). These L mutants are viable in transheterozytes with L-uncovering deficiencies such as Df(2R)ED2354 that carries a mini-white (w^+m) transgene at the proximal breakpoint of the deficiency. (E) A L²-reversion mutant, L^rev6-3, cannot complement L-uncovering deficiencies, resulting in lethality. The mutant heterozygotes show normal or mild eye reduction. Both left and right eyes of an individual are presented to show phenotypic variation. (F–H) New L² eye-reversion alleles, L^D0B (F), L^gR1 (G), and L^P3 (H), could not complement the lethality of L^rev6-3 and L-uncovering deficiencies, suggesting that all these mutations are allelic to L. Heterozygotes of these alleles show normal eyes. (I and J) L-RB-specific lethal mutations. Left and right eyes from an individual show variable phenotypes in L^b5/SM6a (I) and L^b6/SM6a (J). (K and L) L²-revertants from opus[ ]Mohr targeting. All reversion lines including L^A1b (K) and L^A4c (L) are viable in transheterozygotes with L-uncovering deficiency Df(2R)ED2354. These lines lost the L²-specific roo[ ]Mohr during the CRISPR mutagenesis (see text for details). (M) Deduced translations from lethal L mutations (L^D0B, L^gR1, and L^P3) in coding sequences of the last exon. Blue boxes indicate the ZF motifs found in the Drosophila Zfp423 family protein, and orange arrowheads indicate the CRISPR target sites. The location of the L^rev6-3 deletion of six amino acids is also indicated. Dark gray boxes indicate aberrant coding sequences caused by the frameshift mutations prior to stop codons. (N) Nucleotide sequence alignment of L-RB-specific L² eye-reverting mutations, as compared to the progenitor L² or the w¹¹¹⁸ control sequences. The InDel mutations are indicated by red letters or a dashed line, respectively. Blue letters for the coding sequence of L-RB exon 1. Bold letters for polymorphic sequences. Frequencies of the L dominant eye phenotypes shown in the (B−L). are: 60–80% for L¹/SM6a (B), 40–60% for L⁵/SM6a (C), >90% for L²/SM6a (D), 40–80% for L^rev6-3/SM6a (E), >90% for L^D0B/SM6a (F), 100% for L^gR1/SM6a (G) and L^P3/SM6a (H), 40–80% for L^b5/SM6a (I) and L^b6/SM6a (J), and 100% for L^A1c/SM6a (K) and L^A4c/SM6a (L), respectively (n > 1000 for each genotype).

The initial mutagenesis to isolate L²-reversion or loss-of-function alleles of L followed the protocol of direct microinjection of constructed gRNA vectors (U6-DM1 or U6-DM2 in pBFv-U6.2) into L² or wild-type embryos expressing Cas9 in the germline (P[nos-Cas9, y⁺, v⁺]; (Kondo and Ueda 2013) (Figure S1, A and B). L^D0B was isolated by the reversion of the L²/+ eye phenotype, and lethal L alleles (L^gR1 and L^P3) were collected by screening the F1 progeny based on the failure to complement Df(2R)ED2354 that uncovers the L region. Sequence analyses confirmed specific InDel mutations in those loss-of-function alleles (Figure 5M). Mutagenesis by microinjection of gRNA sources yielded very low mutation rates in the L region (<0.1% of >3000 F1 flies screened). Hence, subsequent CRISPR targeting experiments were carried out by using transgenic flies expressing specific gRNA sequences.

The U6-cbm1 transgenic fly line was generated for L-RB-specific mutagenesis using the strategy described by Port et al. (2014, 2015). Target sites are indicated by red arrowheads in Figure 5A. L² or L⁵ mutants carrying the U6-cbm1 transgene were crossed with a ubiquitous Cas9 source (M[Act5C-Cas9.P]ZH-2A or Act-Cas9 in short), and F1 flies (Act-Cas9/w or Y; L²/+; U6-cbm1/+) were examined for reversion of L dominant alleles by somatic targeting (Figure S1C). For germline targeting, L² mutants with the gRNA transgene were crossed with flies carrying a germline source of Cas9 (M[vas-Cas9]ZH-2A, vas-Cas9 in short) to collect L²-reversions by the L-RB-specific targeting (Figure S1D). The mutagenic F1 males carrying both transgenes of Cas9 and U6-cbm1 were collected (vas-Cas9/Y; L²/Sco; U6-cbm1/+), and the F2 progeny was screened for modified eye phenotypes. By following the same protocol, the U6-DE1 transformants that express gRNA sequences flanking opus[ ]Mohr were established to excise the transposon from Mohr strains. Target sites are indicated by green arrowheads in Figure 5A. The gRNA transgene was also used to induce modification of L phenotypes by somatic targeting, or to collect L²-revertants by targeting the opus transposon in the germline of F1 males (Act-Cas9 or vas-Cas9/Y; L²/Sco; U6-DE1/+; Figure S1, C and D).

Male recombination between L² and Oaz-GAL4 insertions by CRISPR targeting

Using preferential homology-dependent repair (HR) of targeted DNA breaks, we designed targeted recombination between the L² mutation and Oaz-GAL4 insertions (P[GawB]NP5288 and P[GawB]GH146, Figure 4B). The gRNA transgenes, U6-cbm1 and U6-csm1 that target distal or proximal sites of Oaz-GAL4 insertions, respectively (Table S1, target sites are indicated by two pairs of red arrowheads in Figure S8A), were used to induce male recombination between L² and Oaz-GAL4 chromosomes. For the targeted recombination, mutagenic F1 males of vas-Cas9/Y; L²/Oaz-GAL4; U6-cbm1 (or U6-csm1)/+ were constructed and were crossed with CyO balancer females (Figure S1E). F2 progeny flies were screened for recombinants that show both dominant traits of parental or modified L²/+ eye phenotypes and mini-white (w^+m) expression by Oaz-GAL4 insertions, and then recombined sequences around target sites were determined (Figure S8).

Figure 4.

Molecular mapping of L mutations. (A) Identification of natural transposons shared by Mohr strains (ft¹, L¹, L², L⁵, and L^r). A genomic map of L/Oaz is adapted from the FlyBase Genome Browser. Shared transposon insertions were found in or nearby the L region: opus[ ]Mohr, roo[ ]Mohr, and a remnant sequence of jockey[ ]Mohr (blue triangles). The dotted line inside opus[ ]Mohr indicates an undefined part of the transposon that could not be amplified by PCR. Note that roo[ ]Mohr is different from roo[ ]828 (gray triangle) described in the FlyBase reference genome. (B) Allele-specific transposons were found in dominant L mutants: roo[ ]L¹, roo[ ]L^r, and hopper[ ]L⁵ (blue triangles). A silencer cis-element, which is the putative target sequence for Polycomb-silencing (Negre et al. 2011), is indicated as a green box. Possible insulator elements (Negre et al. 2010) are indicated as asterisks. Gray triangles indicate P-element insertions in upstream sequences of L-RB, and a red bar indicates the small deletion in L^rev6-3. (C) Identification of L²-specific alterations found within the transposon sequence of roo[ ]Mohr. A 622 bp duplication, a 5 bp deletion, and a nucleotide change are indicated by blue colors.

Immunohistochemistry

An anti-L antiserum was generated by using the C-terminal part of Oaz product (corresponding to G⁷⁶⁸ to the last A¹²²⁸ of L-PB, including eight ZF motifs) that was expressed as a fusion protein with GST, and collected from rats after a series of subcutaneous injections of the purified fusion protein. Immunohistochemistry was performed using the staining protocol described in Patel (1994) with a modification of blocking solution (0.3 M NaCl, 1% NP-40, 0.5% BSA, and 10% NGS) for the more stringent condition. Binding solutions containing the anti-L antiserum were pretreated overnight with fixed w¹¹¹⁸ embryos at 2 hr after egg laying (AEL) before primary staining of sample embryos, or were pretreated with fixed embryos and first instar larvae at 24 hr AEL before staining of larval imaginal discs. Fixed embryos were prepared in the whole-mount preparation procedure described in Patel (1994), and larval tissues including imaginal discs were fixed as described in Spratford and Kumar (2014). Those fixed samples were incubated for 2 hr to overnight at 4° in the aforesaid pretreated binding solutions containing primary antibodies, and were then treated with fluorescent secondary antibodies in a standard binding solution (PBS of 175 mM NaCl, 0.5% Tween-20, and 0.1% BSA) for 1–2 hr at room temperature (RT). After washing over 2 hr at RT and/or overnight at 4° in PBS with 1% Tween-20, samples were mounted in Vectashield with DAPI (Vector Laboratories). Fluorescent images were acquired using ZEISS LSM710 and LSM780 microscope systems.

Data availability

Strains and plasmids are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables. Supplemental material available at figshare: https://doi.org/10.25386/genetics.12618950.

Results

Lobe is independent of CG10109/PRAS40

A P-element insertion (P17) allele was isolated in a previous study as a lethal mutation that enhanced L^si/+ eye phenotype. P17 was an insertion within the CG10109 gene (labeled as P[PZ]PRAS40^P17 in Figure 1B; Chern and Choi 2002). In subsequent genetic analyses, however, P17 could complement a semi-lethal allele L^rev6-3, which was isolated as a revertant suppressor of L² dominant allele. Further tests also indicated that these two mutations are not allelic. While L^rev6-3 was lethal over Df(2R)L48 (Figure 1A), which is a revertant deficiency isolated as a suppressor of L² eye phenotype (Davis and MacIntyre 1988), the lethal P17 mutation complemented the L deficiency as well as L^rev6-3.

To test whether the P17 P-element insertion is responsible for the lethality, this P-element was mobilized to generate revertants by precise excision. However, no viable revertant was found from over 200 independent jumping events. Subsequently, we were able to remove the lethality by backcrossing with an isogenic control strain (Ryder et al. 2004). Viable P17 homozygous flies developed normally without any visible phenotype in the eye (Figure S2B), which is consistent with the report that a knockout allele of CG10109/PRAS40 generated by homologous recombination is fully viable with no detectable developmental defects (Pallares-Cartes et al. 2012). Taken together, these results indicate that the P17 insertion in CG10109 is not lethal. P17 seems to be an amorphic or strong hypomorphic allele of CG10109 since CG10109 protein was nearly undetectable in P17 mutant clones generated in eye discs (Figure S2, C and D). Furthermore, the dominant L²/+ eye phenotypes were not modified by the isogenized P17/+ or PRAS40^KO/+ (Figure S2, J and L), supporting that the L eye phenotype is independent of the P17 insertion.

L^rev6-3 lethality maps to Oaz

Previous studies had mapped the L gene to the 51A2-B1 region by breakpoints of a L-revertant deficiency, Df(2R)L4 (Baker and Ridge 1980). Since L^rev6-3 is a semilethal mutation that suppresses L²/+ dominant eye phenotypes, probably due to an intragenic mutation, we tried to map the recessive lethality of L^rev6-3. To identify the gene for L^rev6-3, we carried out complementation tests between L^rev6-3 and several deficiency chromosomes that uncover a region of 50E–51E (Figure 1A). L^rev6-3, which shows 0–5% survival of homozygous adult flies depending on culture conditions, was clearly lethal when heterozygous for deficiencies such as Df(2R)ED2354, Df(2R)BSC357, Df(2R)BSC668, Df(2R)BSC11, and Df(2R)trix, as well as Df(2R)L48. In contrast, two other deficiencies, Df(2R)BSC700 and Df(2R)Exel8059, complemented the L^rev6-3 lethality (Figure 1A). These tests with deficiencies mapped the L^rev6-3 lethality within a chromosomal region between 51A2 and 51A4, from the distal breakpoint of Df(2R)BSC700 to the proximal breakpoint of Df(2R)Exel8059. In this region, four coding genes have been annotated: Achilles (Achl), phyllopod (phyl), Oaz, and CG10109/PRAS40 (Figure 1B). From this list, we eliminated Achl and phyl because these genes are intact in Df(2R)BSC668, which failed to complement L^rev6-3. Thus, the lethality of L^rev6-3 might map to either Oaz or PRAS40. However, since no sequence difference was detected between L² and its revertant L^rev6-3 in the PRAS40 genomic region, it is likely that the lethality of L^rev6-3 maps to Oaz.

Reduced Oaz suppresses L dominant eye phenotypes

Although L^rev6-3 lethality was mapped to Oaz, it remained to be determined whether the dominant L eye phenotype maps to the same gene. Based on the suppression of L phenotypes by L deficiencies (Baker and Ridge 1980; Davis and MacIntyre 1988), we reasoned that the eye phenotypes of L dominant mutations (L¹, L², L⁵, and L^si; Figure 2) might be suppressed by reduced expression of those four candidate genes located between 51A2 and 51A4 (Achl, phyl, Oaz, and PRAS40, Figure 1B). L¹ and L^si are weak dominant alleles that usually show small dentation in the anterior edge of heterozygous mutant eyes. Comparison of L dominant phenotypes in heterozygous flies indicates that L⁵/+ is much stronger than L¹/+ and L^si/+, but milder than L²/+ which is the strongest L allele. Because L eye phenotypes are highly variable depending on the genetic background, those L mutant alleles were backcrossed for 10 generations to minimize the potential effects of nonspecific genetic background. We then tested whether these L eye phenotypes can be modified by RNAi knockdown of the four candidate genes in developing eye discs using ey-GAL4-driven UAS-dsRNA expression.

First, we checked for genetic interaction between PRAS40 and L mutants. Two RNAi strains against PRAS40, P[GD6869]v38338 and P[GD6869]v38339, could fully restore the small eye phenotype caused by PRAS40 overexpression (hereafter ey > PRAS40). However, no significant modification of dominant L eyes was seen with these PRAS40 RNAi lines. Since both PRAS40 RNAi lines showed off-target lethality when expressed ubiquitously with Act-GAL4 or Tub-GAL4, we generated new transgenic RNAi lines by targeting more unique sequences in the PRAS40 3′ coding region (Materials and Methods). This new PRAS40 RNAi line did not cause any noticeable developmental defects with Act-GAL4, but strongly suppressed the ey > PRAS40 phenotype (Figure S2, E and G), while P17/+ could partially suppress the phenotype (Figure S2, E and F). These results indicate that this new RNAi line efficiently knocked down PRAS40. However, the new PRAS40 RNAi still could not modify L²/+ dominant eye phenotype (Figure 2J).

Next, we tested whether phyl and Achl genetically interact with L mutations. Two independent phyl RNAi strains (P[TRiP.JF03369] and P[GD12579]v35469) failed to modify L phenotypes (Figure 2I). In addition, two lethal deletion mutants of phyl, phyl²²⁴⁵/+ and phyl²³⁶⁶/+ (Chang et al. 1995), did not show any genetic interaction with L^rev6-3 or L dominant mutants. Three P-element insertions have been reported in the promoter region of Achl (Thurmond et al. 2019). One of them, P[GawB]NP4297, was initially lethal and greatly enhanced L eye phenotypes. However, after several generations of backcrossing, this P-element insertion line became viable and was unable to modify L phenotypes. Only one TRiP RNAi strain of Achl (P[TRiP.JF01976]) was available, and this RNAi also did not noticeably modify the L²/+ eye phenotypes (Figure 2H).

Finally, we checked whether Oaz shows any genetic interaction with the dominant L alleles. Strikingly, Oaz RNAi (P[TRiP.GLV21080]) showed strong suppression of the L²/+ eye phenotype to the normal eye and consistently rescued all tested dominant mutant L phenotypes (Figure 2, K, N, P, and R). Two other independent Oaz RNAi lines (P[TRiP.JF01943] and P[KK102438]) also resulted in weaker but significant suppression of L²/+ eye (Figure 2L). Taken together, these data suggest that L dominant eye phenotypes are caused by abnormal upregulation or ectopic expression of Oaz.

To see whether overexpression of Oaz can induce the L mutant phenotype, we used EP lines (P[XP]d09084 and P[GSV1]s-72) that have UAS elements in the upstream of Oaz-RB sequence (Thibault et al. 2004; Kankel et al. 2007; Cruz et al. 2009). When those EP lines were combined with ey-GAL4 to overexpress Oaz in developing eyes, resultant flies showed variable but distinctive eye defects that are similar to that of L dominant alleles (Figure S3, A–D). This further supports the idea that the gain-of-function of Oaz is sufficient to induce the L eye phenotype.

L^rev6-3 mutant has a deletion in the Oaz coding region

As shown earlier, the lethality of L^rev6-3 was mapped to the 51A2-4 region (Figure 1). To identify the molecular lesion of this potential loss-of-function allele of L, we carried out sequence analysis in the 51A2-4 region from Achl to PRAS40, and compared the sequences between L^rev6-3 and its progenitor L². As described above, no sequence difference was found between L² and L^rev6-3 in the PRAS40 genomic region. Genomic sequences around phyl and Achl were also identical between the two mutants. However, a small deletion of six amino acid residues was found in the Oaz coding sequence of L^rev6-3 (Figure 3 and Figure 4B), which corresponds to a putative Mad-binding ZF motif of the Oaz protein (Hata et al. 2000; Krattinger et al. 2007). Since this deletion was found only in L^rev6-3 but not in the progenitor L², this deletion is likely to be the specific L^rev6-3 mutation that suppresses the dominant L² eye phenotype. Since transheterozygotes of L^rev6-3 over L deficiencies uncovering Oaz are lethal, Oaz is an essential gene and L^rev6-3 might be a loss-of-function allele of Oaz.

Based on these genetic and molecular data described above, we conclude that Oaz is the responsible gene for L^rev6-3 mutation and dominant L eye phenotypes. In accordance with the FlyBase guidelines for nomenclature (https://wiki.flybase.org/wiki/FlyBase:Nomenclature) (Thurmond et al. 2019), we propose to rename Oaz as L since L was named earlier based on the mutant phenotypes. Accordingly, Oaz will be referred to hereafter as L.

L dominant alleles contain transposable elements in the L region

Next, we examined genomic sequences from L dominant alleles to identify molecular lesions for the mutant eye phenotypes. All L dominant alleles are spontaneous mutations isolated many decades ago (Lindsley and Zimm 1992; Wang and Huang 2009). When compared with wild-type control sequences of w¹¹¹⁸ or CS, L mutants showed a number of polymorphisms in genomic DNA sequences and in deduced amino acid sequences of the four candidate genes in the 51A2-4 region (Figure 3B). Remarkably, however, all four L alleles (L¹, L², L⁵, and L^r) showed almost identical sequences in the genomic region.

These L dominant mutants were isolated in a similar time period ∼1920 (Lindsley and Zimm 1992). Thus, we looked for a non-L mutant isolated during this time period, assuming that such a mutant might have a similar genetic background as L mutants. Indeed, the ft¹ mutant chromosome isolated by Mohr (Lindsley and Zimm 1992) showed almost identical sequences to the L mutants in the examined genomic region, hence providing control for sequence comparison with the L alleles. Analysis of the genomic region 51A2-4 from ft¹ and the four L dominant mutants (hereafter referred to as the “Mohr strains” after the discoverer’s name) showed interesting sequence differences only in the L region.

The L gene region comprises >30 kb-long sequences within which two different splice forms of L (L-RB and L-RC) have been predicted (FlyBase, GBrowse release 5.57; Figure 4), although only the L-RB form was reported (Krattinger et al. 2007). We confirmed the expression of both forms by RT-PCR from Canton-S wild-type embryos (Figure S4A). Thus, we examined the entire L region including the long intron sequences around L-RC exons to identify molecular lesions in L mutants. The Mohr strains shared insertions of two different transposons around the second exon of L-RC. One was a partial sequence (369 bp) of jockey transposon located just before the exon 2 (jockey[ ]Mohr in Figure 4A), and the other was an expanded form of opus transposon just after the exon 2 (opus[ ]Mohr in Figure 4A). The typical opus transposon is known to be 7.6 kb long. We confirmed the presence of ∼6 kb canonical border sequences of the opus[ ]Mohr, but the more interior region of the transposon could not be amplified by PCR for sequencing, possibly due to additional transposon insertions within the opus element. However, this opus[ ]Mohr was also found in the ft¹ control mutant, suggesting that this transposon is probably not related to the dominant L phenotypes.

Interestingly, among Mohr strains, L¹ has a specific roo transposon insertion in the first L-RB intron, located ∼1.2 kb downstream from the L-RB transcription initiation (roo[ ]L¹ in Figure 4B). Furthermore, another specific roo element was found in the L^r mutant in the second intron of L-RC (roo[ ]L^r in Figure 4B). These roo elements were not detected in the ft¹ sequence, thus the transposon insertions might be the cause of gain-of-function phenotypes of L¹ and L^r mutants. In addition to the roo element insertions, we found that L⁵ has a 1.4 kb hopper transposon in the first intron of L-RC (hopper[ ]L⁵ in Figure 4B).

L² allele shows alterations in the roo[ ]Mohr transposon

The L² mutation causes the strongest eye phenotype among classical L dominant mutations. However, we could not find any molecular difference between L² and the control ft¹ within the sequenced region of L transcription. Hence, we examined whether L² mutation might be located in the 5′ or 3′ intergenic region. We found that all Mohr strains share an intact roo element in the 3′ intergenic region (Figure 4, A and C). This roo insertion is different from roo[ ]828 described in the FlyBase reference genome (GBrowse release 5.57). roo[ ]828 was not detected in all examined stocks including w¹¹¹⁸ and CS controls, as reported in a previous study (Potter and Luo 2010). The roo element found in the Mohr strains, labeled as roo[ ]Mohr, is positioned at 261 bp proximal to the roo[ ]828 insertion, and is ∼1.9 and 5.4 kb away from the L stop codon and the first exon of PRAS40, respectively. This 9088-bp-long roo[ ]Mohr contained a conserved intact long ORF of the roo long terminal repeats (LTR) retrotransposon.

Interestingly, we found that the ORF of roo[ ]Mohr in L² mutant has specific sequence alterations such as a 622 bp duplication, a 5 bp deletion, and a nucleotide change (Figure 4C). Because no other sequence difference was found between L² and the control ft¹ within the genomic region from phyl to PRAS40, the L²-specific roo[ ]Mohr is likely to be the cause of L²/+ eye phenotype that can be suppressed by L knockdown (Figure 2).

New L²-revertants or loss-of-function L alleles fail to complement L^rev6-3 lethality

To substantiate the postulation that the L^rev6-3 small deletion is responsible for the reversion of the L² eye phenotype, we generated new L²-reverting mutations by specific CRISPR mutagenesis that targets the coding sequence of the L gene. The CRISPR target was chosen to cover the coding sequence deleted in L^rev6-3 (Figure 5, A and M, orange arrowheads). L² embryos expressing Cas9 in the germline (L²/CyO, nos-Cas9) were injected with a gRNA construct described in Materials and Methods. From F2 progeny, we isolated a new mutant L^D0B that reverted the L²/+ eye phenotype (Figure 5F). In the isogenized L^rev6-3 heterozygote population (w¹¹¹⁸; L^rev6-3/+), ∼40–80% of flies (n > 1000) showed weak eye phenotypes such as small dents in the anterior part of the eye (Figure 5E) similarly to L¹ (Figure 5B). In contrast, most L^D0B/+ flies had nearly wild-type eyes. Occasionally, L^D0B/+ eyes showed weak L phenotypes but at a significantly lower rate than L^rev6-3/+ (<10%, n > 1000), suggesting that L^DOB is a stronger allele than L^rev6-3 (Figure 5F, as compared to Figure 5E). Molecular analysis of L^D0B indicated that a codon corresponding to A⁶⁰⁰ of L-PB was replaced with insertion of extra eight nucleotides. This change causes a frame-shift that results in an early translation termination after 29 aberrant codons, thus deleting more than half of the C-terminal L product, including 12 out of 21 ZF motifs (Figure 5M). These molecular data are consistent with the stronger reversion of the L²/+ eye phenotype by L^D0B than L^rev6-3. We also confirmed that L^D0B could not complement the lethality of L^rev6-3 and deficiencies uncovering L region, such as Df(2R)ED2354, Df(2R)BSC357, Df(2R)BSC668, Df(2R)BSC11, and Df(2R)L48 (Figure 1A). Hence, L^DOB mutation is allelic to L^rev6-3.

L revertant mutations like L^rev6-3 were made in the L² mutant background. To generate specific L loss-of-function mutations without L² dominant mutation, we carried out an additional CRISPR mutagenesis in the wild-type background. By the same CRISPR strategy used for isolating L^D0B, we targeted a further upstream site marked by the left arrowhead in Figure 5M. We obtained two lethal alleles, L^gR1 and L^P3 (Figure 5, G and H), that show 10- and 11-bp deletion in the last exon of L, respectively. These mutations result in frame-shifts after A²⁶⁸ and A²⁶⁹ of the L-PB product, respectively, thus removing most part of Zn-finger motifs of L protein (Figure 5M). Both of these mutations are allelic to L as they failed to complement the lethality of L²-revertant mutations (L^rev6-3 and L^D0B) and deficiencies uncovering the L gene. Thus, L²-reverting mutations such as L^rev6-3 and L^D0B are loss-of-function mutations in L, different from classical L gain-of-function alleles that are viable as transheterozygotes with L deficiencies (Figure 5, B–D). Our data also support that the reversion of the L²/+ eye phenotype by L^rev6-3 is attributed to the small deletion in the coding region of L.

It has been reported that L has a role in the development of the posterior spiracle in the embryo (Krattinger et al. 2007). By L-GAL4 expression (P[GawB]NP5288 and P[GawB]GH146 in Figure 4B), we found L expression in embryonic or larval posterior spiracles (Figure S7C and S9A). We also checked whether L^rev6-3 and newly isolated lethal L mutants exhibit any defects in the posterior spiracle development. The majority of L^gR1 homozygotes could not complete embryogenesis, although 0–5% of embryos hatched and died as first instar larvae. In contrast to L^gR1/+ heterozygotes that show the normal pattern of posterior spiracles, homozygous larvae showed reduced and malformed posterior structures (Figure S5B). Similar defects were also found in L^P3 homozygous larvae or in transheterozygotes of L^rev6-3 with L^gR1 or L^P3 mutations (Figure S5, C–G). Consistent with L-GAL4 expression in posterior spiracles, these data suggest that all new L mutations affect posterior spiracle development.

L-RB-specific mutations suppress L²/+ eye phenotype and causes lethality

As described above, we confirmed that two splice forms of the L gene, L-RB and L-RC, are expressed (Figure S4A). We also showed that ectopic expression of L-RB by two EP elements (P[XP]d9084 and P[GSV1]s-72, Figure 4B) inserted upstream to the L-RB transcription can cause dominant L eye-like phenotypes (Figure S3, A–D). To verify the causal relationship between the L eye phenotype and ectopic expression of L-RB, we targeted L-RB-specific sequences of the L-RB exon 1 (Figure 5A) using the CRISPR-Cas9 method in L dominant alleles.

Transgenic flies expressing the specific gRNA sequences were crossed with L² or L⁵ allele expressing Cas9 under the ubiquitous Act5C promoter (Act-Cas9). This somatic CRISPR targeting induced a wide range of suppression of the L²/+ eye phenotype due to different levels of L-RB mutant cells in the mosaic eyes (Figure S6, L²/+, Act-Cas9 > U6-cbm1). These data from the somatic targeting indicate that the L²/+ eye phenotype is dependent on the level of L-RB function. However, L⁵/+ eyes were scarcely affected in the same somatic CRISPR targeting. Because L⁵/+ eyes were restored successfully by RNAi targeted to common sequences shared by both L splice forms (Figure 2P), the lack of L-RB targeting effects on L⁵/+ supports that the L⁵ dominant phenotype is largely dependent on the L-RC form rather than L-RB, which is consistent with the hopper[ ]L⁵ insertion near the first exon of L-RC.

We also generated germline mutations in L-RB as suppressors of L² from a genetic screening of the L-RB-specific CRISPR targeting (Figure 5, I and J). L^b5 showed a 29-bp deletion comprising the 3′-end of L-RB exon 1 and the 5′-splicing donor sequence of intron 1. As for L^b6, a sequence substitution in the CRISPR target site resulted in a frame-shift from the 24th codon of L-PB followed by early translation stop (Figure 5N). Both L^b5 and L^b6 mutations suppressed the L²/+ eye phenotype, resulting in a partial rescue like the hypomorphic L^rev6-3 allele (Figure 5, E, I, and J). L^b5 and L^b6 could not complement all L loss-of-function mutations tested including L^rev6-3 and L^D0B, indicating that the L-RB splice form is necessary for normal development.

CRISPR mutagenesis in the L region involves homology-directed repair of targeted DNA breaks

We have sequenced the entire L region except for the opus[ ]Mohr transposon inserted near the L-RC exon 2. As described above, opus[ ]Mohr is shared by Mohr strains including classical L dominant alleles and the control ft¹ (Figure 4A). Therefore, although we found alterations in the 3′ intergenic roo[ ]Mohr of L² mutant, it was necessary to test whether opus[ ]Mohr in L² mutant has no change from that of the ft¹ control strain. However, we could determine only ∼6 kb terminal regions from both ends of opus[ ]Mohr probably because more interior part of the transposon could not be amplified by PCR, perhaps due to a structural change in opus[ ]Mohr such as insertion of extraneous repetitive sequences.

As an alternative approach, we attempted to excise the opus[ ]Mohr transposon by CRISPR mutagenesis and see whether the L²/+ eye phenotype can be reverted by removing the transposon. To excise opus[ ]Mohr, we used U6-DE1 transgene expressing gRNA to target two sites flanking the insertion in the intron 2 of L-RC: one at 58 bp before the proximal opus LTR and the other at 292 bp after the distal LTR of the transposon (target sites as green arrowheads in Figure 5A, see Materials and Methods). Somatic targeting of the opus[ ]Mohr with ubiquitous Cas9 expression showed variegated suppression of the L²/+ phenotype as seen in the previous targeting of L-RB exon 1 (Figure S6, L²/+, Act-Cas9 > U6-DE1). This suppression seemed to be specific to L² since L⁵/+ eyes were rarely modified by the same somatic targeting.

We then isolated 22 independent L²-revertant lines from the germline CRISPR mutagenesis. All of these lines showed complete reversion of the L²/+ eye phenotype (Figure 5, K and L). Intriguingly, these lines could also complement the lethality of L loss-of-function mutants, such as L^D0B, L^gR1, and Df(2R)ED2354, indicating that targeted removal of opus[ ]Mohr did not impair the function of L. These L²-revertants were analyzed at the molecular level to determine whether the reversion was due to loss of opus[ ]Mohr. In about half of the revertants (12 of 22 lines), the opus[ ]Mohr transposon was either completely excised or mostly deleted except an opus LTR at the end. Conversely, the other half of reversion lines (10 of 22 lines) showed no difference from the original L² sequence or only minor changes in sequenced amplifiable regions of the transposon. Hence, the insertion of opus[ ]Mohr transposon may not be responsible for the L²/+ eye phenotype, raising a possibility that reversion of the L²/+ eye phenotype might be due to a different change in the L region during opus[ ]Mohr targeting.

In an attempt to identify such change, we further analyzed L region in L²-revertant lines by using genomic polymorphisms between L² and the other homologous chromosome. Remarkably, analysis of polymorphic markers in L²-revertant lines revealed frequent recombination between L² and L⁺ homologous chromosomes around target sites. In all 22 reversion lines, the 5′ intergenic region consists of the L² chromosomal sequences whereas the 3′ intergenic region was derived from the homologous Sco-marked L⁺ chromosome. This indicates that two parental chromosomes were recombined around the CRISPR target sites during the mutagenesis, which resulted in the loss of the L²-specific roo[ ]Mohr. Since the loss of roo[ ]Mohr in the 3′ intergenic region coincides with complete loss of L²/+ phenotype, the L²-specific alteration in the roo[ ]Mohr transposon appears to be a major cause of the L²/+ eye phenotype.

Our data above suggest that CRISPR mutagenesis is accompanied by homology-dependent repair (HR) of the targeted DNA break and/or homologous recombination in the male germline expressing Cas9. We further tested the HR-dependency of CRISPR targeting events by using deficiency chromosomes that cover or uncover the targeted L region. Chromosomal deficiencies that uncover L (Df(2R)ED2354, Df(2R)BSC357, and Df(2R)BSC668, Figure 1A) and cover the targeted region (Df(2R)BSC700 and Df(2R)Exel8059, Figure 1A) were made transheterozygous with L² in the presence of vas-Cas9 > U6-DE1 or vas-Cas9 > U6-cbm1 (mutagenic F1 males genotype: vas-Cas9/Y; L²/Df(2R)X; U6-L-gRNA/TM2). As expected, no L²-revertant F2 progeny was found from CRISPR targeting when L² was transheterozygous with deficiencies that uncover L. In contrast, dozens of L²-revertants were collected when L² was heterozygous with deficiencies that cover L (Table S2). Further, we could also detect frequent phenotypic recombinations between L² eye and mini-white (w^+m) eye color marker associated with Df(2R)BSC700 (proximal to L). Next, we tested another homologous chromosome, PRAS40^EY10689, which also possesses CRISPR target sites in the L region. From the mutagenic transheterozygotes (L²/PRAS40^EY10689), we could collect L²-revertants. However, no phenotypic recombinant of L² and w^+m eye was found because both the L²-specific roo[ ]Mohr and the PRAS40 P-element w^+m marker are located distal to the CRISPR target sites. This suggests that the CRISPR target sites had been recombination breakpoints between the distal L² and the proximal w^+m of Df(2R)BSC700 (Table S2).

Altogether, these results suggest that L² suppression by CRISPR targeting at opus[ ]Mohr was due to loss of the L²-specific roo[ ]Mohr through the HR pathway or meiotic recombination between L² and its homologous chromosome in mutagenic F1 males.

L protein is expressed ectopically in L²/+ eye disc

Our data so far suggest that L dominant eye phenotypes are likely due to abnormal regulation of L in eye discs. To identify the misexpression of L protein, we generated anti-L antiserum by using a 461 aa C-terminal L protein fragment containing eight Zn-finger motifs. In wild-type, anti-L immunostaining was not detected in eye disc (Figure 6A and Figure S3G), although there was a strong expression in other tissues including embryonic and larval brain, ventral ganglion, anterior and posterior spiracles, and leg disc (Figure S7). In addition to the reported L function in the posterior spiracle development (Krattinger et al. 2007), we found that L RNAi or mutation causes loss of adult claws (Figure S7, E–G), indicating that L expression in the distal region of leg disc (Figure S7D) is important for claw development. We also examined the pattern of L transcription by using L-GAL4 insertion lines (P[GawB]NP5288 and P[GawB]GH146, Figure 4B) as reporters. In wild-type, L-GAL4 expression was also undetectable in eye disc (Figure 6D), whereas it was strongly expressed in posterior spiracles and larval/adult brains (Figure S7C and Figure S9).

Figure 6.

Ectopic L expression in L mutant eye discs. (A–C) Anti-L antiserum staining in early third instar larval eye discs. No L immunostaining was detectable in eye-antenna imaginal discs of control stocks (A). Ectopic L expression was detected in L²/+ mutant eye discs (B) and L^rev6-3/+ eye discs (C) from the late second to early third instar larvae. (D–F) Ectopic GAL4 activities in L²/+ mutant eye discs. L-GAL4 L² recombinants were used to detect ectopic L-GAL4 (P[GawB]GH146 or P[GawB]NP5288) activities in third instar eye discs. Real-time L-GAL4 expression and clonal expression were visualized by RFP and GFP, respectively, using the G-TRACE system. Control L-GAL4 without recombination with L² showed no GAL4 expression in eye discs (D–D”). Ectopic GAL4 expression was clearly detected in L^GDb1 (P[GawB]GH146 + L²) recombinant discs (E–E”) and L^5Db5 (P[GawB]NP5288 + L²) recombinant disc (F–F”). Two L-GAL4s, P[GawB]GH146 and P[GawB]NP5288, show similar ectopic GAL4 activities. Bar, 100 μm. Full genotypes for (D–F) are as follows: (D) y¹ w*/ w*; P[GawB]GH146/ P[UAS-RFP], P[UAS-FLP], P[UAS>>GFP], (E) y¹ w*/ w*; L^GDb1/ P[UAS-RFP], P[UAS-FLP], P[UAS>>GFP], and (F) y¹ w*/ w*; L^5Db5/ P[UAS-RFP], P[UAS-FLP], P[UAS>>GFP].

Next, we examined whether L protein is expressed ectopically in L²/+ or L^rev6-3/+ mutant eye discs. Since the reduction of eye discs can be clearly detected already in the second instar larvae, we examined L²/+ eye discs from the larval stage. Anti-L staining was not detected in early second instar discs. However, we found ectopic L expression in L²/+ eye discs from the late second- to early third-instar larvae (Figure 6B). Ectopic L expression was enriched mainly along the dorsoventral (DV) midline of the late second-instar eye disc. In early third-instar eye discs where retinal differentiation has been initiated, ectopic L expression was detected along the DV boundary in the undifferentiated anterior region (Figure 6B). However, this expression pattern was not maintained in later third-instar stages, probably due to the preferential loss of ventral tissues (Singh et al. 2006).

We also examined eye discs from L^rev6-3/+ larvae. Since L^rev6-3 is an inframe small deletion (6 aa) mutation that partially suppresses the L²/+ eye phenotype (Figure 3A), we reasoned that the mutant may express the mutated L protein in nearly normal shaped eye discs. Indeed, we could confirm the DV midline pattern of ectopic L expression in normal shaped L^rev6-3/+ eye discs (Figure 6C), although the expression level was weaker than that of L²/+ eye discs.

L-GAL4s are expressed ectopically in eye disc by the L²-specific roo transposon

The observation of ectopic L protein in L²/+ eye discs suggests that L² mutation may cause the misregulation of L gene expression. To test whether L² mutation leads to transcriptional misregulation of L, we used intrinsic reporter activities of L-GAL4 insertions (P[GawB]GH146 and P[GawB]NP5288). The expression of these L-GAL4 reporters was not affected by L mutations in trans position. Therefore, we decided to generate recombinants between L-GAL4 insertions and L²-specific roo[ ]Mohr transposon and test whether L-GAL4 activity can be induced ectopically by L² in cis position. Since L-GAL4 and roo[ ]Mohr are <9 kb apart, it would be difficult to generate recombinants by conventional meiotic recombination. Hence, we utilized our finding that the HR-dependent CRISPR targeting in the germline can trigger male recombination around the target site. We designed a CRISPR mutagenesis to induce recombination between L-GAL4 insertions and L²–specific roo[ ]Mohr. CRISPR target sequences were chosen in the regions proximal or distal to L-GAL4 insertions. Proximal and distal guide RNA was encoded by U6-csm1 or U6-cbm1 gRNA construct, respectively, as shown by red arrowheads in Figure S8A. Targeting was induced in F1 males of L²/L-GAL4, vas-Cas9 > U6-cbm1 (or U6-csm1). Interestingly, only the distal targeting by cbm1 transgene could generate L-GAL4 L² recombinants that show red eye color with L² eye phenotypes (either parental L²-like or partially suppressed). Sequence analysis of these recombinant lines confirmed the presence of both L-GAL4 and roo[ ]Mohr. Among those, recombinants showing strong L² suppression had frame-shifting deletions in the targeted L-RB exon 1. Conversely, other recombinants showing weak or no suppression had inframe changes in the L gene (Figure S8A), consistent with the data that L² phenotype is suppressed by reduced L function (Figure 2, Figure 3, and Figure 5).

To detect ectopic L-GAL4 expression in these cis-recombinants carrying the GAL4 driver and L² mutation, we examined GAL4 activities in eye discs by using the G-TRACE system (Evans et al. 2009) that allows both real-time expression and clonal expression of L-GAL4. Unexpectedly, upon crossing with the G-TRACE reporter, the heterozygous L²-recombinants (L-GAL4 L²*/UAS-RFP, UAS-FLP, Ubi>>GFP) showed abnormally delayed larval growth, resulting in few adult survivors for an unknown reason. In contrast, no developmental defect was noticed in the progeny from control crosses between the G-TRACE reporter and the original L-GAL4 insertions or L². Normally, L-GAL4 expression is prominent in larval posterior spiracles and larval/adult brains (Figure S9). In these tissues, the real-time (L-GAL4 > RFP) and clonal reporter expression (L-GAL4 > FLP, Ubi>>GFP) in L-GAL4 L² recombinants were indistinguishable from the control L-GAL4 reporter expression pattern of nonrecombinants. Reporter expression patterns were similar in different recombinant lines showing L² suppression.

As expected, neither real-time nor clonal reporter expression was detected in wild-type eye discs carrying the control L-GAL4 insertions (P[GawB]GH146 or P[GawB]NP5288) (Figure 6D). However, we detected ectopic GAL4 activity in eye discs from L-GAL4 L² recombinants (Figure 6, E and F). The real-time L-GAL4 activity (L-GAL4 > RFP) was enriched along the DV boundary region of eye disc similar to the pattern of ectopic L protein expression in L²/+ and L^rev6-3/+ mutant eye discs (Figure 6, B and C). However, the clonal reporter expression (L-GAL4 > FLP, Ubi>>GFP) was much broader to cover most part of the eye-antennal disc, suggesting that L is expressed more broadly in L²/+ eye discs at early stages. Such ectopic reporter expression was consistently found in different L²-recombinants like L^GDb1 (P[GawB]GH146 recombined with L²) or L^5Db5 (P[GawB]NP5288 recombined with L²), suggesting that ectopic expression is induced by the L² mutation. Despite the ectopic L reporter expression, these eye discs showed nearly normal eye shape since functional L-PB protein is not expressed due to the frameshift mutations in L^GDb1 and L^5Db5.

Taken together, these data suggest that L² dominant eye phenotype is due to misexpression of L in eye disc induced by the altered form of roo[ ]Mohr.

Discussion

We have presented genetic and molecular evidence that Oaz is the L gene. Our data show that L dominant eye phenotypes can be restored by reducing the level of L and that L expression is ectopically induced in L²/+ mutant eye disc. These results suggest that classical L dominant alleles are gain-of-function mutations that inhibit eye development by causing L misexpression.

Role of transposable elements in L dominant phenotypes

Our data reveal that all classical L dominant mutants tested are gain-of-function mutants associated with the insertion of natural transposons in the L region. All Mohr strains including the control ft¹ share two retrotransposons, opus[ ]Mohr near L-RC exon 2 and roo[ ]Mohr in the 3′ intergenic region of L, as well as a fragment of jockey element (Figure 4). In addition to these common insertions, each of L dominant mutants carry an additional specific transposon such as hopper[ ]L⁵, roo[ ]L¹, or roo[ ]L^r, which were not found in the ft¹ control strain. These allele-specific transposons were inserted in different introns of the L transcription unit. Besides, we could also identify the L²-specific mutations within the retroviral sequence of roo[ ]Mohr, although this transposon is shared by all Mohr strains. The L² mutation found in the 3′ intergenic region of L could induce ectopic GAL4 expression in eye disc when the L-GAL4 insertions were in cis position with the mutation.

Among natural transposons found in the L region, hopper is a member of the terminal inverted repeats (TIR)-dependent DNA transposon family, and opus and roo elements are LTR family retrotransposons. These natural transposons are frequently found in different Drosophila strains (Kaminker et al. 2002; Kapitonov and Jurka 2003; Linheiro and Bergman 2012), and there have been several reports of spontaneous mutations associated with transposon invasions. The sex-lethal allele Sxl^M4 is the only known mutation caused by hopper insertion (Bernstein et al. 1995). Hence, L⁵ seems to be the second example of hopper-induced mutation. Multiple opus insertions have been associated with developmental mutations such as opus[ ]N^fa-1, opus[ ]lz³⁴, and several tube mutations (Gorman and Girton 1992; Crew et al. 1997; Clemmons and Wasserman 2013), all of which are loss-of-function mutations. Dozens of roo retrotransposon-induced mutations have been reported, including a gain-of-function mutation, Gla (Brunner et al. 1999).

At least three possible mechanisms can be considered for how the insertion of the natural transposons in the L region can cause L dominant eye phenotypes. First, L might be misexpressed by regulators of the inserted transposons, as seen in Gla mutation. The Gla eye phenotype was proposed to result from wg misexpression in pupal eyes, driven by the roo LTR sequence inserted in the wg promoter region (Brunner et al. 1999). Similarly, a Drop mutation (Dr^Mio) with a roo family-related retrotransposon inhibits eye development by ectopic expression of the Dr gene that encodes an Msh homeobox transcription factor (Mozer 2001). Gla and Dr^Mio mutations by roo or roo-related retrotransposons affect late larval or pupal stage eye development. On the other hand, L mutant phenotypes can be detected in earlier larval stages prior to retinal differentiation.

Second, L mutant eyes may result from derepression of L silencing by transposon invasions. It is intriguing that putative Polycomb (Pc)-response element and tandemly aligned insulator elements have been identified in the L region (Nègre et al. 2010, 2011; Roy et al. 2010). Genome-wide analyses also indicated that the L region is highly enriched with chromatin domains for the Pc-mediated repression (Schwartz et al. 2006; Filion et al. 2010; Kharchenko et al. 2011). Interestingly, transposons associated with L dominant phenotypes are located near the insulator motifs or the Pc silencer (Figure 4B). Hence, L dominant eye phenotypes might be caused by altered epigenetic silencing of L expression by transposon insertions or by alterations within the existing retroviral element.

Third, L⁵ mutant has an insertion of hopper[ ]L⁵ near the first exon of the L-RC isoform, far from the insulator cluster (Figure 3B). L-RB form-specific CRISPR targeting strongly suppresses L² mutation but has much weaker effects on L⁵. In contrast, the L⁵/+ phenotype is effectively suppressed by RNAi targeting both L-RB and RC forms. This suggests that the L⁵/+ phenotype depends more on the L-RC splice form than the RB form. The N-terminal region of vertebrate Zfp423 family proteins contains a nuclear remodeling and histone deacetylation (NuRD) motif implicated in transcriptional repression of target genes including itself (Lin et al. 2004; Hong et al. 2005; Bond et al. 2008; Matsubara et al. 2009; Hesse et al. 2010). This motif is present in L-PC but not L-PB (Figure S4B). It remains to be studied whether the N-terminal motif of L-PC is also involved in transcriptional repression and whether L-RC is aberrantly transcribed by the hopper[ ]L⁵ insertion.

Homology-dependent repair of CRISPR targeting-induced DNA breaks

We used a variety of CRISPR targeting methods to characterize L dominant eye phenotypes. One of them was to revert the L² dominant phenotype by targeted excision of the opus[ ]Mohr. In this process, we found that CRISPR targeting in the male germline results in frequent meiotic recombination at or around the target sites. Indeed, our data suggest that L² suppression by the opus[ ]Mohr targeting was due to meiotic exchanges between L² mutation and wild-type homologous chromosome rather than deletion of the transposon per se. However, when the same CRISPR targeting was made in L² transheterozygous with chromosomes deficient in the L region, L² suppression was not found, supporting its HR-dependency.

Compared to HR, the nonhomologous end-joining pathway (NHEJ) has been reported to be a fast and efficient DSB repair response, and mutagenesis by CRISPR targeting in Drosophila has been attributed largely to the error-prone NHEJ repair in the germline (Bier et al. 2018). However, the choice between competing DSB repair pathways is affected by multiple regulatory mechanisms depending on cell cycle or cellular contexts (Mao et al. 2008; Chapman et al. 2012; Zaboikin et al. 2017). For example, an analysis of several meiosis mutations revealed that NHEJ is negatively regulated in Drosophila oogenesis (Joyce et al. 2012). Our data also suggest that germline repair of induced DSBs during our CRISPR mutagenesis at L follows the HR pathway predominantly over NHEJ.

Using the targeted recombination strategy, we could successfully generate L-GAL4 L² recombinants in cis position. These recombinant GAL4 reporters helped us demonstrate ectopic L expression by the L² mutation in developing eye disc. Ectopic expression of L-GAL4 reporter was also confirmed by the expression of L protein in L²/+ mutant eye discs. These data suggest that the L² eye phenotype is due to ectopic L expression in eye disc and provide an explanation for the suppression of all L dominant phenotypes by L RNAi or mutations.

Our results from G-TRACE experiments show that ectopic real-time expression of L-GAL4 reporter is preferentially induced in the DV boundary region of the early third instar eye disc, while clonal L-GAL4 expression is widespread in most part of the eye disc. Hence, L-GAL4 seems to be expressed broadly in the early eye disc, but only the equatorial region remains to be active while dorsal and ventral expression is repressed by early third instar stage. As described above, all identified classical L mutations are associated with the insertion of natural transposons or alterations in the existing transposable element. It is possible that ectopic induction of L by transposable elements might antagonize critical genes involved in early eye development. The DV boundary pattern shows a similarity to that of eyegone (eyg)—a PAX-family gene important for early eye development (Jang et al. 2003; Dominguez et al. 2004; Yao and Sun 2005). Indeed, a reduced reporter expression of eyg was reported in a weak L dominant allele, L^fee (Wang and Huang 2009). Consistently, we had previously shown a phenotypic enhancement of another weak L eye, L^si, by reduction of N signaling (Chern and Choi 2002) that is required to maintain eyg expression in the growing eye primordium (Dominguez et al. 2004; Wang et al. 2008). It was also proposed that Eyg has the second function to repress wg expression that inhibits the initiation of retinal differentiation in eye disc (Yao and Sun 2005; Ekas et al. 2006). It is noteworthy that the L²/+ eye phenotype is suppressed by reducing Wg signaling (Singh et al. 2005, 2006). Molecular mechanisms for transposon-induced misregulation of L and affected regulators of eye development in different L mutants remain to be studied.

Identification of a molecular lesion for loss-of-function L^rev6-3 mutation

In the analysis of the L^rev6-3 mutation, we found a deletion of six amino acid residues in the 10th ZF motif located within the putative Mad-binding domain of L. Among the tandem repeats of ZF motifs, the central ZFs containing the putative Mad-binding domain is well conserved between the Drosophila L and its vertebrate homologs (Hata et al. 2000; Krattinger et al. 2007). Since the short deletion was uniquely found in L^rev6-3, this mutation is likely to be responsible for the intragenic suppression of the L² dominant eye phenotype. Because homozygous L^rev6-3 mutation is semi-lethal, the putative MAD-binding domain containing the 10th ZF motif seems to be critical for normal functions of L during development. The vertebrate Zfp423 family proteins interact with diverse transcription effectors, including SMAD proteins, O/E factors, retinoic acid receptors, the Notch intracellular domain, and HDACs (Tsai and Reed 1997; Hata et al. 2000; Huang et al. 2009; Hesse et al. 2010; Masserdotti et al. 2010). Among these, the BMP signal-induced binding to SMADs or the transcriptional regulation of an inhibitory SMAD is important for the differentiation of various tissues (Ku et al. 2006; Gupta et al. 2010; Kamiya et al. 2011; Addison et al. 2014). Identification of the L^rev6-3 mutation in the putative Mad-binding domain raises a possibility that L may also interact with Dpp signaling during development. In addition, our data suggest that the SMAD binding site of Zfp423 family proteins plays an important role in invertebrates as well as vertebrates.