We first designed a synthetic gene circuit where the successful GFP expression was coupled to the activity of an intron-encoded gRNA (Figure 1A). Specifically, this circuit was designed so that both the splicing of the gRNA-containing intron (located within the reporter gene) and the efficient expression of the gRNA cassette itself were necessary to achieve high levels of eGFP expressions. For this purpose, we used a set of three gRNAs that had previously been described to be capable of recruiting a dCas9 transcriptional activator to the ×5 QUAS sequence (Lin et al., 2015). In this plasmid-based system, we provided the dCas9-VPR-T2A-mCherry activator together with a reporter plasmid containing the 5xQUAS motif upstream of a minimal Hsp70 promoter and the eGFP cassette which in turn hosted the intron constructs to be tested. We assessed two intron variants based on either the Drosophila melanogaster ftz intron or a synthetic intron previously characterized in a mammalian cell culture (Kiani et al., 2014). The QUAS-targeting gRNAs were incorporated into the introns upstream of the branch point, as summarized in Figure 1B. We either included the full U6:3 promoter and a T6 terminator or simply the gRNA sequence itself without any additional regulatory sequences. In the former case, we expected that the gRNA expression would be driven independently from the expression of the GFP gene whereas in the latter case, the spliced intronic RNA would itself act as a gRNA, as has been shown before (Figure 1B). A third variant was also generated where the gRNA was located directly on the 5′ strand of the poly-pyrimidine stretch where the leading T nucleotides were meant to act to terminate transcription.

The reporter and activation constructs were co-transfected into Drosophila melanogaster S2 cells to evaluate the performance of gRNA constructs and to simultaneously enable splicing and the transactivation of the eGFP expression. GFP fluorescence was quantified to determine the relative efficacy of each intron-encoded gRNA variant. As a negative control, we employed the ftz and synthetic intron base constructs prior to the insertion of any gRNA cassette. As a positive/activation control, we used these base constructs together with the dCas9-VPR activator into which we cloned a cassette co-expressing transactivating gRNA1. As expected, we detected, for both intron variants, low levels of the eGFP expression in the negative controls (Figure 1C). The positive control yielded a strong fluorescent signal relative to the negative control (8.3-fold and 71.3-fold induction in the ftz and synthetic constructs, respectively), suggesting that transactivation was successful and that the intron control constructs lacking gRNAs were splicing efficiently. When analyzing combinations of gRNAs and the two intron variants, we found that no significant fluorescent induction was triggered by the constructs containing each gRNA alone (Figure 1C). This suggested that either splicing or the level of gRNAs liberated as spliced RNA was insufficient in these constructs. When we supplied gRNA1 for transactivation by co-transfecting the activation control plasmid, the ftz and synthetic minimal constructs carrying gRNA1 did yield a fluorescent signal comparable to that of the respective U6-gRNA1-T6 constructs (Supplementary Figure S1). This suggested that the minimal constructs did not express sufficient gRNA for sufficient transactivation but could splice efficiently when the gRNA was provided in trans. The constructs featuring the U6 promoter and terminator sequences consistently performed best in this assay for both intron types. The U6-gRNA1-T6 construct yielded a fluorescent signal 10.4-fold (ftz intron) and 137.5-fold (synthetic intron) higher than the controls. The U6-gRNA1-T6 construct was also the only construct performing significantly better than the activation control (1.9-fold and p < 0.0001), although several constructs yielded an overall higher mean level of fluorescence. Only a single experiment (U6-gRNA1 in conjunction with the synthetic intron) featuring gRNA-encoded upstream of the poly-pyrimidine stretch showed a significant level of induction (Figure 1C). Overall, we concluded that full gRNA expression cassettes can be located within small intron sequences, allowing both efficient gRNA expressions and splicing, and we took these designs forward for evaluation in transgenic flies.

For the generation of autonomous integral gene drives, we chose two Drosophila melanogaster target loci, bam and rcd-1r, the regulatory regions of which had previously been utilized for the successful generation of gene drives in the fly (Chan et al., 2011; Chan et al., 2013; Simoni et al., 2014). These two genes were also chosen because in both cases, we were able to identify Cas9-targetable sites and PAM motif close to the start codons that would allow the insertion of the Cas9 coding sequence upstream of the coding sequence of these genes (Figure 2A). Bam (CG10422) has been associated with the fusome, a germ cell–specific organelle, and it contributes to the fate determination of germline stem cells in males and females with loss of function, leading to the production of aberrant sperm and eggs. The bam promoter has been frequently used to drive a germline-specific expression in D. melanogaster. Rcd-1r (CG9573) is a retrogene (Bai et al., 2007) that has been implicated in male fertility with loss-of-function mutants found to be semi-sterile or sterile (Chen et al., 2012). It shows strong expressions in adult testes, and although rcd-1r 5′UTR has been assayed mainly for inducing male-specific gene drives, rcd-1r itself is also expressed during early embryogenesis, specifically in zygotic germ cell nuclei starting from stage 8 and in early developing gonads (Tomancak et al., 2002; Lecuyer et al., 2007; Tomancak et al., 2007; Hammonds et al., 2013; Wilk et al., 2016). We next designed integration constructs where intron-encoded gRNAs targeting bam or rcd-1r were located within the Cas9 open reading frame which was designed to link to the host locus via a T2A signal (Figure 2A). We chose the ftz intron which is derived from Drosophila, and the full U6:3-gRNA-T6 design had previously shown good performance in the cell assay. The constructs also featured a removable 3xP3-CFP transformation marker flanked by loxP sites and 5′ and 3′ regions of homology to the bam or rcd-1r loci. In both cases, the transgenes were successfully integrated into these two germline loci, which were confirmed by PCR genotyping of G1 individuals which were outcrossed to balancer strains. Following intercrosses, we observed that homozygous bam ^d+CFP male and female individuals were sterile, whereas we managed to successfully establish a homozygous strain rcd-1r ^d+CFP which showed no obvious fertility defects. In order to remove the 1.6 kb transformation marker cassette flanked by loxP sites which we expected to interfere with correct splicing and host gene expression, we next crossed homozygous rcd-1r ^d+CFP or TM6B-balanced bam^d+CFP heterozygotes to a Cre-recombinase expression strain. The transhemizygote progeny was intercrossed, and their CFP-negative progeny was then propagated in single crosses, and their offspring in turn was screened molecularly for successful excision of the fluorescent marker cassette. Following this strategy, we successfully established a homozygous strain rcd-1r ^d. In the case of bam insertion, however, we did not recover any offspring from 10 independent single crosses of markerless individuals. This suggests that although the transgene (now consisting of only the Cas9 coding sequence and the intron-encoded gRNA) was designed to not interfere with the function of the host gene, the presence of the insert did not allow for the sufficient or correct expression of bam and was causing sterility.

To investigate the gene drive of the constructs established at the two loci, we conducted a series of homing assays (Figure 2B). We crossed male or virgin female rcd-1r ^d hemizygotes, rcd-1r ^d+CFP hemizygotes, and bam ^d+CFP hemizygotes with wild-type (w ¹¹¹⁸ ) flies and determined the transmission of the transgene in the progeny. This was either carried out by tracking the fluorescent CFP marker or, in the case of the markerless rcd-1r ^d insertion, by making use of molecular genotyping. Between 18 and 30, single crosses of one male and three females were established for each condition, and crosses that failed to produce offspring were discarded. Figure 2B summarizes the results of these experiments. We found no significant deviation from Mendelian transmission rates for the bam ^d+CFP transgene (51.3 and 49.2% transgenic progeny for male and female crosses, respectively). This suggests that Cas9 and/or the gRNA were not correctly expressed in these constructs in all likelihood due to the presence of the CFP marker. Given these results and due to the observed sterility of homozygotes, for the remaining experiments, we focused on the rcd-1r insertions alone. We did observe a significant gene drive in rcd-1r ^d+CFP male (62.8% transgenics, p < 0.001) but not in the female (51.4%) cross. High rates of transmission were observed for the rcd-1r ^d construct from both hemizygous males (77.1%, p < 0.001) and females (80.5%, p < 0.001). We concluded that although some levels of the expression of Cas9 occur in the presence of the CFP marker gene in rcd-1r ^d males, more substantial rates of homing were enabled only once the marker cassette had been removed. The rcd-1r gene had previously been used for a male-specific gene drive; here however, we observed the drive in both sexes. Together with the established expression pattern of rcd-1r in early embryogenesis, this suggested that the observed gene drive could also be zygotic. To distinguish between the gene drive in the germline and zygotic gene drive, we crossed homozygous rcd-1r ^d /rcd-1r ^d males or females with the wild type. In the progeny, we did not find any evidence of somatic homozygosity in 92 genotyped individuals, all of which were rcd-1r ^d /+ heterozygotes, suggesting thus that the rcd-1r ^d gene drive is confined to the germline.

In order to examine target site modifications in individuals who failed to inherit the drive construct, wild-type progenies were taken from the rcd-1r ^d,CFP homing cross in pools of 50 individuals. Because we outcrossed transgenic males with wild-type females, this experiment allowed us to understand the variety of indels formed in the absence of selection to retain rcd-1r function. Wild-type sized alleles were amplified from a subset of these crosses (n = 12) and compared against pools of w ¹¹¹⁸ individuals that represent the pre-existing variation at the rcd-1r locus in the laboratory colony (n = 3). Resulting reads were analyzed using CRISPResso2 (Clement et al., 2019), and the indel formation rate was determined (Figure 2C). Reads were pooled across all biological replicates and analyzed for frequency of substitutions, deletions, and insertions proximal to the rcd-1r target site. We found no strong evidence of pre-existing target site variations at the rdc-1r locus in the w ¹¹¹⁸ background, but 12.1% of reads from the homing cross were found to carry indels. We found that the majority of indels were predicted to affect the rcd1-r coding sequence, but interestingly, less than the predicted 2/3 of indel events were predicted to result in a translational frameshift (32.4%). We attributed this discrepancy to the presence of a common indel variant found in 7/12 replicates and representing a total of 11% of all detected variant reads. It is likely generated by an ATGAGT microhomology with a predicted MMHEJ score of 240.8 (Bae et al., 2014) and results in a 24-bp deletion which, although it maintains the rcd-1r reading frame, leads to the elimination of the original start codon and its context (Supplementary Table S1).

In order to determine whether insertion of the IGD drive component affected the fitness of the transgenic rcd-1r strains, phenotypic assays were performed to measure basic life history traits related to fertility and fecundity. We quantified the egg output and subsequent hatching rates of homozygous rcd-1r ^d and rcd-1r ^d,CFP strains and the w ¹¹¹⁸ controls. Individual homozygote females and males were crossed in vials and allowed to mate for 24 h. At least eight replicates were performed per strain, with assays performed over 72 h, flipping mated females into a fresh vial every 24 h. The egg output was quantified per female, per day for each transgenic strain (Figure 2D). The vials were maintained at 25°C in order to quantify the rate of eclosion. Three days after the eclosion of pupae started, the number of adults in each vial was counted, and the hatching rate was calculated (Figure 2D). No statistically significant difference was observed between the rcd-1r ^d and rcd-1r ^d,CFP drive lines, when compared to a w ¹¹¹⁸ control line in terms of fertility or hatching rate. Given that rcd-1r had previously been implicated in male fertility, this suggested that at this locus, the IDG components, even in the presence of the full fluorescent marker cassette, did not interfere with fly fertility.

We next performed population cage experiments to determine the dynamic behavior of the rcd-1r ^d integral gene drive in caged Drosophila populations of 200 individuals over 10 fly generations. We performed this experiment in four replicate populations and two release conditions where hemizygous rcd-1r ^d males represented 15% or 25% of the starting male populations. This corresponds to 7.5 and 12.5% starting allele frequencies of the rcd-1r ^d drive allele, respectively. Because rcd-1r ^d is a markerless gene drive, we used molecular genotyping to determine the presence of hemizygous and homozygous transgenic individuals at generations 1, 3, 5, 7, and 10. We compared the observed dynamic to that predicted by a stochastic agent-based model which was parameterized with the fitness and homing rates for rcd-1r ^d . We found that as predicted by the model which assumes that maintaining the rcd-1r function is necessary for male but not female fertility, the rcd-1r ^d drive spreads rapidly but reaches a maximum allele frequency of approximately 80% with a similar peak frequency under both release conditions. These frequencies are slightly less than those predicted by the models and could be due to the effects that we did not account for computationally, for example, low levels of maternal Cas9 deposition into the embryo. This experiment validates the results of the individual genetic crosses and shows a successful drive of a minimal integral gene drive at the population level. We performed Sanger sequencing of non-transgenic individuals to determine target site variants that had accumulated in these cage populations by generation 10. Out of a total of 40 samples, 13 carried a 6-bp deletion which was detected in six out of the eight cage populations (Figure 3B, Supplementary Table S1). This suggests that selection for maintaining the rcd-1r function leads to the rise of a prominent variant featuring a two-amino acid deletion at the protein’s N-terminus.

Experiments were performed in Schneider 2 cells (Thermo Fisher Scientific, United Kingdom). Cells were cultured at 25°C in an ambient atmosphere. Cells were maintained in a complete growth medium composed from 90% Schneider’s Drosophila medium (Thermo Fisher Scientific, United Kingdom), and 10% FBS (Sigma-Aldrich, United Kingdom). Cells were maintained in T-25 flasks, with 0.2-μM filter caps. Visualization was performed using a Nikon TMS inverted microscope and a ×10 confocal lens. Cells were passaged biweekly, at 1:4 ratios, at which point they had achieved a density of >1 × 10⁶ cells/ml. The cell count was performed using a Scepter 2.0 Cell Counter. Confirmation of the mycoplasma-free status was performed by PCR. Constructs used for experimental work in S2 cells were prepared for transfection by being midiprepped using E. coli TOP10 cells. Following transformation, individual colonies were selected and used to inoculate 5 ml of LB-Miller starter cultures, supplemented with ampicillin. Starter cultures were incubated for 8 h at 37°C in a shaking incubator at 220 rpm. They were then used to inoculate 500-ml falcon flasks containing 100 ml of LB-Miller broth, supplemented with ampicillin. Falcon flasks were incubated overnight at 37°C in a shaking incubator at 220 rpm.

Annotated DNA plasmid sequences have been provided in Supplementary Material S1. Briefly, sequences for the constructs used for the cell-based assay were derived from pQUAST (Addgene plasmid #24349) by the insertion of gene-synthesized fragments and validated by Sanger sequencing. To generate U6:3-gRNA-T6 gRNA cassettes for the generation of gene drive constructs, spacers (bam GTC​GGG​CAA​CGA​CGA​CCA​GCA​GT, AAA​CAC​TGC​TGG​TCG​TCG​TTG​CCC; and rcd-1r GTC​GAT​GAG​TGC​GGA​ACC​AAG​TC, AAA​CGA​CTT​GGT​TCC​GCA​CTC​AT) were synthesized as two complementary ssDNA oligos (Eurofins Genomics, Ebersberg, Germany), with BbsI compatible overhang sites in the pCFD3-dU6:3gRNA (Addgene plasmid #49410) (Port et al., 2014), which were then cloned into the Bsp119I site within pBS-Hsp70-SpCas9-ftz, followed by the subsequent insertion of the 3xP3:CFP:SV40 cassette. For each locus, 1-kb homology arms were synthesized (ATUM, CA, United States), and BbsI or BaeI sites were introduced to produce a linearized backbone into which the drive cassette was inserted by the Gibson assembly.

Fluorescence in S2 cells was analyzed by using a three-laser BD LSRFortessa analyzer. Live, single cells were selected by gating forward and side scatter. For each condition analyzed, at least 500 gated events were recorded, with three replicates per condition. The following voltages were used—forward scatter (FSC): 170 V, side scatter (SSC): 253 V, 530-nm channel: 390 V, and 610-nm channel: 591 V. Excitation of eGFP was performed using a 488-nm blue laser and by using a 530/30 bandpass filter. The excitation of mCherry was performed using a 561-nm yellow laser and by using a 610/20 bandpass filter. Transfection of single color constructs was used for the construction of compensation matrices to account for spectral overlap.

Embryos were dechorionated in 50% bleach, prior to microinjection and aligned on 10 × 10 mm coverslips. The needles were pulled using a P1000 Sutter micropipette puller (Sutter, United Kingdom). Injections were conducted with w ¹¹¹⁸ embryos at 500 ng/μl, consisting of a 400 ng/μl drive construct and 100 ng/μl pnos-Cas9 helper plasmid. Following microinjection, embryos were retained on coverslips and placed onto 100 mm × 100 mm agar plates. Upon hatching, first instar larvae were placed into a yeast paste and transferred to plugged vials to mature at 22°C. All stocks were maintained at 18°C, while all experiments were conducted at 25°C under Arthropod Containment Guidelines Level 2 conditions.

Rcd-1r drive lines were generated by performing 10 single crosses of rcd-1r ^d,CFP with y1 w ^67c23 ; sna ^Sco /CyO, and Pw [+mC] = CrewDH1 virgins to excise the 3xP3:CFP:SV40 reporter cassette located within the Cas9-located ftz intron of the drive component. G1 progenies were all markerless and were intercrossed to produce homozygous rcd-1r ^d G2 progenies. Excision of the CFP reporter cassette was confirmed by Sanger sequencing.

The oviposition and hatching assay (Simoni et al., 2014) was performed using the w ¹¹¹⁸ line as a control. Virgin females and males were collected and aged for 24 h. Single crosses were subsequently performed, with mating being allowed to proceed for 48 h. Males were subsequently removed, and females were allowed to lay eggs in vials containing standard yeast food for three consecutive days, with flies transferred into a new vial every 24 h. At least eight replicates were performed per line, with the number of eggs per female per day quantified. The number of eclosed adults arising from said eggs was then quantified, with at least eight replicates observed per line.

Hemizygotes were obtained by crossing single rcd-1r and rcd-1r ^d,CFP males to w ¹¹¹⁸ virgins. In the case of bam balanced w ¹¹¹⁸ , bam/TM6B and Tb [1] males were crossed with w ¹¹¹⁸ virgins. Subsequent G1 heterozygotic progenies were screened such that only w ¹¹¹⁸ bam individuals were selected to act as parents for homing assay crosses. Hemizygous progenies were collected such that the females were virgins, and all individuals were 1–3 days old. Single males were crossed with three females in each cross. Male and female hemizygous transgenics were assayed separately to determine their sex-specific effects. In cases where replicates failed to yield offspring, they were discarded and where lines carried a CFP reporter cassette, the gene drive carrying the progeny was identified using an Olympus MVX10 fluorescent microscope for fluorescence. A markerless rcd-1r progeny was genotyped by extracting genomic DNA using Chelex 100 resin. An extraction solution was produced by suspending 1.25 g of Chelex resin in 100 ml of RNase-free water supplemented with 4 ml of proteinase K solution (20 mg/ml). This solution was added in 100 μl aliquots to each well containing a fly. Plates were sealed using adhesive plate foils. Incubation occurred in an Eppendorf Thermomixer-C, for 2 h at 55°C and 700 rpm, and then, proteinase K was inactivated by incubation for 20 min at 99°C. Plates were stored at −20°C. PCR reactions using primers TAG​CAA​AGT​CAG​GGC​GTA​GC and CAC​CGG​GAT​AAG​CCC​ATC​AG using REDTaq DNA polymerase were conducted as follows: 98°C for 2 min, followed by 35 cycles of 98°C denaturation step for 20 s, a 59°C annealing step for 30 s, and a 72°C extension step for 30 s, followed by a final 72°C extension step for 7 min, and then a 4°C hold. PCR reactions were visualized using 1% agarose gels, and the presence of a 609-bp band was considered to be a positive indication of a transgenic individual.

Amplicon sequencing was performed on a CFP-negative progeny of the rcd-1r ^d,CFP homing assay. In total, 50 CFP-negative G2 progenies were selected at random from a given biological replicate, and pooled genomic DNA extraction was performed for 12 replicates. Then, 200 control w ¹¹¹⁸ flies of a commensurate age were selected, and pooled genomic DNA extraction was performed in three replicates. Qubit quantified yields were normalized to 50 ng/μl using Tris-EDTA. PCR was then performed using primers ACA​CTC​TTT​CCC​TAC​ACG​ACG​CTC​TTC​CGA​TCT​GAC​GCT​TTT​CCA​CAG​CAT​GG and GAC​TGG​AGT​TCA​GAC​GTG​TGC​TCT​TCC​GAT​CTT​TCG​GTC​CTT​TCT​CGC​TTG​A producing an amplicon of or around 320 bp. Primers included a partial scaffold compatible with Illumina NGS sequencing. PCR conditions were as follows: 98°C denaturation step for 2 min, followed by 23 cycles of 98°C for 10 s, 57°C for 10 s, and 72°C for 30 s, followed by a final 72°C extension for 7 min, and a 4°C hold. Successful amplification of PCR products was confirmed by loading 5 μl of each sample onto a 1% agarose gel, running at 100 V for 60 min, and visualizing. PCR products were subsequently purified using a QIAGEN MinElute PCR Purification kit (QIAGEN, CA, United States), and purified DNA products were eluted in 10 μl Tris-EDTA. Purified DNA was quantified using a Qubit fluorometer as before and normalized to 20 ng/μl, in at least 20 μl of the total volume using Tris-EDTA. Samples were processed by Genewiz (Genewiz, Leipzig, Germany), and the data were processed using CRISPResso2 (Clement et al., 2019).

Experiments were performed in a 250-ml conical bottle, plugged with cellulose acetate plugs (VWR, United Kingdom) and filled with 50 ml of standard yeast-based fly food (12lt water, 240 gr polenta, 90 gr agar, 960 gr fructose, 1,200 gr brewer’s yeast, 60 ml Nipagin 15% w/v in ethanol, and 90 ml propionic acid). Flies were allowed to mate for 5 days, before being recovered using CO₂, and stored at −80°C for subsequent analysis. Three days after G1 individuals had eclosed, the adults were retrieved from each bottle and counted. Then, 200 individuals were selected at random and placed into a fresh bottle to begin the next generation. Any excess flies were frozen at −80°C for later analysis. This cycle was repeated until generation 10. To genotype DNA, extraction was performed as described, and multiplex PCR reactions were setup in 96-well plates, corresponding to each gDNA plate. Primers AGA​AGC​TGT​CGT​CCA​CCT​TG, ACG​TGC​TTT​CGG​TCC​TTT​CT, and AGG​TGT​TCT​TGC​TCA​GCT​CC that bind to the transgene and the endogenous locus generate a product of 586 bp when the rcd-1r ^d allele is present and a 292-bp product for the wild-type allele. We used the REDTaq DNA polymerase mastermix (VWR, United Kingdom) as follows: denaturation at 98°C for 2 min, followed by 35 cycles of the 98°C denaturation step for 20 s, 59°C annealing step for 30 s, and the 72°C extension step for 40 s, followed by a final 72°C extension step for 7 min, and then a 4°C hold. Plates were sealed, as before, with an adhesive foil. A total of 10 rcd-1r ^d hemizygous G10 samples from each replica of both release conditions were further analyzed through Sanger sequencing in order to determine the presence of target site variants.

We employed a simple object-oriented stochastic agent-based discrete generation model written in C#. The model allowed to fully account for the genetic interactions of autonomous and non-autonomous CRISPR-based drive elements and the generation and tracking of R1 and R2 resistance alleles at all modeled loci. Briefly, we considered populations of 200 individuals where WT, R1, R2, and drive constituted the possible genotypes at the rcd-1r locus. We included the experimentally observed drive parameters and a 95% loss of male fertility for a loss of the rcd-1r function (R2 homozygotes). In each generation, all adults were randomly mated, and 200 offspring were randomly selected for constituting the next generation.