Exploiting a Y chromosome-linked Cas9 for sex selection and gene drive

Development of a Y chromosome-linked Cas9 strain

Despite the availability of numerous transgenic fly lines in D. melanogaster, including multiple Cas9−expressing lines, there is a complete lack of transgenic lines expressing Cas9 protein from the Y chromosome. To engineer flies expressing Cas9 from the Y chromosome, we generated three plasmids containing SpCas9 driven by the vasa promoter which is expressed in both the ♂ and ♀ germline and soma¹⁸. As the Y chromosome is notoriously gene-poor and consists of silenced repetitive DNA²⁷, we incorporated a marker to assess promoter activity as done previously^18,28. Downstream of the vasa promoter we included a SpCas9−T2A-eGFP cassette which encodes SpCas9 as well as a self-cleaving T2A peptide and an eGFP-coding sequence. We also incorporated a tdTomato transformation marker driven by the eye-specific 3xP3 promoter and flanked by gypsy and CTCF insulators to improve overall expression levels by acting as barrier elements that can block the propagation of heterochromatic structures into adjacent euchromatin²⁹. These components were surrounded by homology arms to aid Y chromosome insertion into three distinct locations via CRISPR-mediated HDR (Only SGyA HDR is shown in Fig. 1a and b)²⁵. To facilitate insertion, these transgenes were injected, along with an in vitro transcribed sgRNA complexed with recombinant SpCas9 protein into a fly line encoding a genetic source of Nanos-SpCas9³⁰. Unfortunately, no transformants were obtained for SGyB and SGyC constructs (Fig. 1c). Of the 540 embryos injected for SGyA, 53% survived, and subsequent outcrossing to w[1118] (WT) resulted in a single transformant G1 ♂ (Fig. 1c, d). To expand this line, and remove the genetic source of Nanos-spCas9, this G1 ♂ was outcrossed to WT ♀’s. In agreement with the paternal inheritance of the Y chromosome, we observed 100% of the G2 transgenic ♂’s, with no ♀’s, expressing the tdTomato eye marker (Fig. 1d). A stock containing the Y chromosome-linked Cas9 transgene was established, and from hereon is referred to as SGyA, displaying exclusive ♂-specific inheritance patterns which continued into all subsequent generations. We molecularly verified the presence of the transgene by PCR and Sanger sequencing across the transgene-insertion junctions using genomic DNA of SGyA ♂’s (Fig. 1e). In establishing the SGyA stock, we observed a range of variable eye maker expressions in adult ♂’s that ranged from moderate to undetectable fluorescence, presumably resulting from heterochromatic silencing effects (Supplementary Figure 1A). Despite this range in marker expression, weak and moderate marker expressing males still displayed sufficient Cas9 activity that was capable of editing and generating high rates of lethality in males that inherited a sgRNA transgene targeting PolG2 (Supplementary Figure 1C). We, therefore, maintained the SGyA line by allowing the ♂s with varied fluorescent marker expression to mate with WT ♀s each generation (Supplementary Figure 1B).

a The Cas9 transgene design for Y chromosome insertion via CRISPR/Cas9−mediated cleavage and homology-directed repair (HDR). Homology arms (HA) flank two insulator sequences, GypSy and CTCF, and a vasa-controlled Cas9−T2A-eGFP. An eye-specific marker (Tdtomato) allows for the identification of transgenic flies carrying the transgene. For microinjection, a source of Cas9 and gRNA were provided to cleave the Y chromosome. The SGyA template was provided and inserted into the Y chromosome through HDR. b Karyotype of transgenic SGyA males and non-transgenic females. c The number of embryos injected for the SGyA, SGyB, and SGyC transgenes, survival to the larval stage of injected embryos, and rate (no. of independent transgenic individuals found/no. of embryos injected) is shown. No transformants were obtained for SGyB and SGyC. d White light and fluorescent images of the SGyA line compared to WT. There is an overall faint yet distinguishable expression of the fluorescent marker in SGyA individuals. Please refer to Fig. S1 to see the range of fluorescence in the SGyA eye marker. e PCR confirmation of transgenic male flies harboring the SGyA transgene. Primers corresponding to the left and right genomic insertion regions were used to amplify both sides to ensure the transgene was present. Amplification was performed at least twice by two independent scientists. The expected band size for the left side primer pair is 1.690 kb. The expected band size for the right side primer pair is 1.893 kb.

Given that the SGyA line encodes a T2A-eGFP marker, similar to an available autosomal linked Vasa-Cas9−T2A-eGFP line², we were able to visually compare expression levels between these two strains. To do this, 3–4 day old ♂ testes (structure depicted in Supplementary Figure 2A) were dissected and imaged to assess relative eGFP expression levels comparing WT (negative control) (Supplementary Figure 2B, B’), autosomal linked Cas9 (Supplementary Figure 2C, C’), and SGyA (Supplementary Figure 2D, D’). We expected eGFP expression to manifest in the testes (long structures curled around seminal vesicles; Supplementary Figure 2) due to the role of vasa in germline development³¹. As expected, no eGFP was detected in WT testes (Supplementary Figure 2B’), but visible eGFP fluorescence was present in both autosomal Vasa-Cas9 and SGyA testes and seminal vesicles indicating robust expression of the transgenes in the ♂ germline (Supplementary Figure 2C’ and Supplementary Figure 2D’, respectively).

Quantification of Cas9 expression

To quantify the expression of the Y-linked vasa-Cas9, we performed RNA-sequencing from 3–4 day old ♂’s, using WT ♂’s and autosomal linked vasa-Cas9 as negative and positive controls, respectively. We detected robust expression of the dsRed and eGFP markers and Cas9 in autosomal samples, with comparatively lower expression of tdTomato, eGFP, and Cas9 from SGyA samples, with no significant expression observed in control WT samples (Supplementary Data File 1). A DeSeq analysis revealed that autosomal Cas9 transgenic flies have 5858 differentially expressed genes (Supplementary Figure 3A) and SGyA transgenic flies have 476 differentially expressed genes (Supplementary Figure 3B, Supplementary Data File 2 and Supplementary Data File 3, respectively). Of the list of differentially expressed genes in both samples, 321 genes were found to be the same between samples. To get a sense of what genes are differentially expressed, we also included a Gene Ontology analysis (GO) in our DeSeq output. The top upregulated genes in the autosomal Cas9 vs WT dataset (log2FoldChange) include the DsRed and eGFP markers, Cas9, several long non-coding RNAs, genes associated with defense response, cuticle development, proteolysis, and Hsp70 (Heat-shock proteins) among others (Supplementary Data File 2). In the SGyA vs WT dataset, top upregulated genes include Cas9, GFP, and tdTomato markers, several long non-coding RNAs, defense response genes, and proteolysis genes (Supplementary Data File 3). Taken together, these data confirmed expression from the Y-linked vasa-Cas9, albeit it was slightly weaker than the autosomal linked vasa-Cas9.

Gene editing using the Y-linked vasa-Cas9

To genetically assess the efficacy of the SGyA line, we crossed SGyA ♂’s to ♀’s from strains encoding gRNAs targeting genes that result in clear visual phenotypes when disrupted. For example, we used an available strain simultaneously expressing multiplexed sgRNAs targeting four genes including sepia, ebony, curled, and forked, each flanked by tRNA’s (tRNA-sgRNA) from a single promoter³² (Fig. 2a, b). We also tested five additional strains encoding sgRNAs targeting wingless, cut, apterous, twisted, and scalloped (Supplementary Figure 5)³³. To compare the effects of the Y chromosome linkage on Cas9 activity we used the autosomal vasa-Cas9 as a positive control¹⁸. In crosses with a genetic source of Cas9, mutant phenotypes were seen in the F1 generation, whereas no phenotypes were observed in WT crosses lacking Cas9 (Fig. 2c and Supplementary Figure 4A, Supplementary Figure 5). In experimental crosses involving SGyA ♂’s and the tRNA-sgRNA, we found that ♀ F1 progeny did not inherit the Cas9 transgene, and therefore did not display mutant phenotypes as expected. F1 ♂’s exhibited subtle mosaic mutant phenotypes in three out of the four target genes, showing that activity of the Y-linked Cas9 is specific to ♂’s (Fig. 2c and Supplementary Figure 4B; Table 1). No mutants were recovered for the cu target and were therefore excluded from the figure. A greater proportion of the mutant ♂’s were single mutants for the forked gene. Typically, forked null mutants tend to have several bristles that are short and have split ends. In the forked mutants, we saw a few bristles that were short, and even fewer bristles had split ends. In the case of ebony, null mutants have a dark cuticle, however, in our mutant ♂’s, we observed mosaic patches of ebony cuticle on the thorax (Supplementary Figure 4B). The inheritance of the SGyA transgene was PCR confirmed in F1 ♂’s for all crosses (Supplementary Figure 6A). In comparison, an autosomal source of Cas9 produced F1 ♂’s and ♀’s with increased penetrance and expressivity of mutant phenotypes; typically producing triple mutants (ebony-forked-sepia mutants ~92%). (Fig. 2c and Supplementary Figure 4C; Table 1). For example, F1 progeny from autosomal crosses had several mosaic patches of the dark cuticle as compared to SGyA mutants. In addition, mutant phenotypes were seen for sepia, whereas none were seen for SGyA progeny. Interestingly, F1 progeny maintained a wild-type allele in the curled gene regardless of the source of Cas9, suggesting there was reduced cleavage at this target site. The sequences derived from the F1 progeny in the autosomal crosses revealed indels in the genes ebony, and sepia, (Supplementary Figure 7). For the progeny derived from SGyA crosses, sequenced target sites showed the wild-type alleles with multiple peaks suggesting somatic mosaicism in the individuals. Further subcloning of the PCR amplicons and sequence analysis showed separated indels for ebony but not curled, forked, or sepia. Crosses between homozygous sgRNA ♀’s and WT ♂’s showed no mutant phenotypes as expected (Supplementary Figure 5B–F, 5B’–F’).

a The multiplexed tRNA-gRNA transgene was used to determine the functional capacity of a Y-linked Cas9 to cleave four phenotypic genes, sepia (se), ebony (e), curled (cu), and forked (f). Flanking the gRNA’s with tRNAs enables expression from a single promoter and processing of the multiplexed gRNAs. Karyotype on the right depicts the location of the four target sites. No cu mutants were obtained and were thus omitted. b Crossing schematic of experiment. Homozygous tRNA-gRNA females were outcrossed to either WT, SGyA, or autosomal Cas9 males. F1 progeny were expected to have either no phenotype and/or a range of mutant phenotypes. Those expected to have a phenotype are indicated in blue dashed borders. c Percentages of F1 progenies with single mutations, double mutations, or triple mutations. F1 progeny from genetic crosses involving SGyA demonstrated single, double, and triple mutants with only 38% of the progeny being composed of f mutants. Crosses involving an autosomal source of Cas9 produced mainly triple mutants. For experimental crosses, seven replicate crosses were set up. For the control, only five replicate crosses were set up. A two-way ANOVA with Tukey’s multiple comparisons was performed on the total mutants (both male and female) comparing wild type and SGyA or autosomal Cas9 data to determine significance. Error bars in black represent the mean ±SEM. ****p < 0.0001. Source data is provided as a Source Data file.

In crosses with ♂s with an autosomal source of Cas9 and ♀ with the single U6:sgRNAs, both F1 ♂ and ♀ progeny displayed phenotypes. For example, when twisted is targeted, both ♂’s and ♀’s have twisted abdomens (Supplementary Figure 5B”). We observed embryo/early larvae and pupae lethality when cut and wingless were targeted (Fig. 5c” and S5D”, respectively). When apterous (ap) and scalloped (sd) were targeted, both ♂’s and ♀’s were affected (Supplementary Figure S5E, F; Supplementary Figure 5E”, S5F”). However, in ap, a small proportion of ♂’s and ♀’s retained WT phenotypes (Supplementary Figure 5E”). In experimental crosses involving SGyA ♂’s and the single U6:sgRNA expression system, we found that ♀ F1 progeny did not display mutant phenotypes as expected. Sequencing the target sites from these females showed no indels (Supplementary Figure 8). Crosses between SGyA ♂’s and sgRNA ♀’s targeting twisted (tw) produced ♂’s with disfigured abdomens which affected the numbers of F1 ♂’s emerging (Supplementary Figure 5B, 5B”’) compared with crosses with WT ♂’s lacking a source of Cas9. Crosses from the SGyA and lines expressing gRNAs for cut (ct) and wingless (wg) caused ♂ lethality at the embryo/early larvae and pupae, respectively, and produced exclusively adult F1 ♀’s (Supplementary Figure 5C, D; Supplementary Figure 5C”’, 5D”’). Crosses with lines expressing gRNAs targeting ap and sd produced wing deformities that prevented ♂’s from emerging from the puparium, affecting the final number of adults counted (Supplementary Figure 5E, F; Supplementary Figure 5E”’, F”’). Taken together, these data suggest our SGyA line is able to efficiently produce mutant phenotypes in a ♂ specific manner.

Efficient sex selection by exploiting sex chromosome-linked Cas9

Given the efficiency of Lethal Biallelic Mosaicism (LBM)^2,3,18,26, we hypothesized that this mechanism could be exploited, in combination with sex chromosome-linked Cas9 elements, as a novel method for sex selection that we term SEx LinkEd CRISPR selecTion (SELECT). To explore this hypothesis, we opted to target an essential haplosufficient gene, DNA Polymerase gamma subunit 2 (PolG2, DNA polymerase γ 35 kDa, CG33650) required for the replication and repair of mitochondrial DNA³⁴. Importantly, high levels of biallelic somatic mosaicism of PolG2 are lethal²⁶. To compare cleavage efficiencies, we outcrossed either WT, SGyA, autosomal Cas9, or X-linked Cas9 ♂’s to a 2nd chromosome-linked strain expressing a sgRNA targeting PolG2 (referred to as U6.3-gRNA#1^PolG2) driven by the U6.3 promoter².

When WT ♂’s are outcrossed to heterozygous U6.3-gRNA#1^PolG2 ♀’s, four expected F1 phenotypes are observed; meanwhile, homozygous U6.3-gRNA#1^PolG2 ♀’s produce only two expected phenotypes and the surviving F1 individuals correspond with the genotypes expected from these crosses (Fig. 3a, b; Table 2). When autosomal vasa-Cas9 ♂’s are crossed with heterozygous U6.3-gRNA#1^PolG2 ♀’s, all transheterozygous F1 progeny perish, while F1 ♀’s (w−;CyO+;Cas9+) and ♂’s (w+;CyO+;Cas9+) expressing Cas9 but lacking the gRNA transgene were recovered (Fig. 3c; Table 2). The crosses involving homozygous U6.3-gRNA#1^PolG2 ♀’s, and ♂’s as a source of autosomal Cas9 produced 100% lethality of all F1 regardless of sex, since all the progeny inherited the U6.3-gRNA#1^PolG2 from the ♀’s and the Cas9 from the ♂’s and were subjected to LBM (Fig. 3d; Table 2). We next crossed ♂’s from X-linked Nanos Cas9 (Bloomington Fly Stock line # 54591) to heterozygous U6.3-gRNA#1^PolG2 ♀’s, and all transheterozygous ♀’s died. From this cross, we recovered only F1 ♂’s lacking Cas9 (w-;CyO+;Cas9− and w+;U6.3-gRNA#1^PolG2/CyO−;Cas9−); and F1 ♀’s (w+;CyO+;Cas9+) expressing Cas9 but lacking the gRNA transgene (Fig. 3e). The crosses involving homozygous U6.3-gRNA#1^PolG2 ♀’s produced 100% ♂’s since all F1 ♀’s inherited the X chromosome-Cas9 from the ♂’s and the gRNA from the U6.3-gRNA#1^PolG2 ♀’s (Fig. 3f; Table 3). Finally, when SGyA ♂’s are crossed with heterozygous U6.3-gRNA#1^PolG2 ♀’s, only ♀’s (w-;CyO+;Cas9− and w+;U6.3-gRNA#1^PolG2/CyO−;Cas9−), and F1 ♂’s (w+;CyO+;Cas9+) lacking the gRNA transgene were recovered, since all transheterozygous F1 ♂’s died (Fig. 3g; Table 3). Similarly, the crosses involving homozygous U6.3-gRNA#1^PolG2 ♀’s with SGyA ♂’s produced ~98.6% viable ♀’s (w+;U6.3-gRNA#1^PolG2/CyO−;Cas9−) and 1.4% ♂’s (Fig. 3h; Table 3). Taken together, these results demonstrate a novel and efficient genetic sexing technique by exploiting LBM (Fig. 3i) using sex-linked Cas9 lines crossed to homozygous gRNA lines targetting essential genes.

In a, c, e, g Heterozygous U6.3-gRNA#1^PolG2 females were crossed to a male that was either WT, SGyA, or autosomal Cas9. Below the cross is the outcome of the F1 progeny. Each F1 fly with a corresponding genotype is associated with either X’, X”, X”’, or X”” labels. The gRNA transgene is marked with a non-fluorescent orange-red eye maker. a No lethal phenotypes were seen in negative control crosses. c In autosomal Cas9 and gRNA crosses, both F1 transheterozygous males and females were not recovered due to the lethal effects of PolG2 cleavage. e In X-linked Cas9 crosses, both F1 transheterozygous males and females were not recovered. Only individuals without gRNA survived. g In SGyA crosses, only F1 transheterozygous males inheriting both SGyA and the U6.3-gRNA#1^PolG2 transgene resulted in lethality (as indicated with red “X”). Females survived. b, d, f, h Homozygous U6.3-gRNA#1^PolG2 females were outcrossed to a male that was either WT, SGyA, or autosomal Cas9. b No lethal phenotypes were seen in negative control crosses. d No surviving offspring were recovered in autosomal Cas9 crosses. f All F1 U6.3-gRNA#1^PolG2 males survived and F1 females perished in X-linked-Cas9 crosses. h All F1 transheterozygous males resulted in lethality while F1 females survived in SGyA outcrosses. i Mechanism and schematic depicting lethal mosaicism in progeny from paternal vs maternal inheritance of Cas9. Lethality is only observed in males which inherit Y-linked Cas9. However, all progeny perish when inheriting maternal Cas9 due to maternally deposited Cas9. Red shaded boxes in Punnett square represent lethality. Mosaic skulls represent lethal mosaicism. Instances, where this symbol is seen, represents no progeny was recovered. Blue and pink bars represent the percentage of males and females, respectively. Black bars represent the standard error of the mean(SEM). Ten replicate crosses were set up per experiment. A two-tailed unpaired student’s t test is used to determine the significance of percentages compared with WT. (refer to Table 2 for ANOVA comparisons) ****p < 0.0001; **p < 0.005; *p < 0.05. N.s. no significance. Source data is provided as a Source Data file.

Characterization of the Y-encoded vasa-Cas9 fly line as a split gene drive

To validate the utility of SGyA in a split gene drive context (Table 3), we genetically crossed the SGyA transgenic flies with a previously generated HomeR gene drive element (GDe)²⁶. The GDe was composed of a re-coded polymerase gamma subunit 2 rescue (PolG2, DNA polymerase Ɣ 35-kDa, CG33650), a PolG2 gRNA, and a marker (3xp3-GFP) that enables scoring GDe inheritance. This transgene should permit the survival of flies inheriting the GDe and the lethality of flies that harbor biallelic NHEJ events. To assess its functionality as part of a sex-biased split drive, we performed genetic crosses (in pentaplicate), by crossing SGyA ♂’s to GDe ♀’s (Fig. 4a). Similar crosses with an autosomal source of Cas9 were also performed for comparison (Fig. 4b). Negative control was performed using WT ♂’s outcrossed to GDe ♀’s (Fig. 4c). Before assessing the inheritance rate of the GDe, we determined the hatching rate of embryos produced from crosses between GDe and WT, GDe and SGyA, and GDe and autosomal Cas9. We found no significant differences between the hatching rate of WT and SGyA (unpaired t test, p value = 0.1836), however, autosomal Cas9 crosses often produced fewer larvae compared to the control (unpaired t test, p value = 0.0094) (Fig. 4d).

a The crossing schematic involves the paternal Cas9 from the Y chromosome. The Cas9 transgene is only passed through the male germline. A SGyA male is crossed to homozygous GDe females to produce transheterozygous males and GDe-only females. The F1 transheterozygous male is then outcrossed to a w- females to assess GDe inheritance in F2 progeny. b The crossing schematic involves the paternal Cas9 from an autosome. A homozygous male harboring Cas9 on an autosome is outcrossed to a homozygous GDe female. Transheterozygous F1 males are subsequently outcrossed to w- females to assess GDe inheritance. Both sexes are affected. c Crossing schematic of the negative control cross. d Percent of F2 eggs hatched in crosses involving the control, transheterozygous SGyA males and transheterozygous autosomal Cas9 males. e Inheritance of the GDe in F2 progeny among different sources of Cas9. f Inheritance of the GDe in F2 and F3 progeny from crosses involving a transheterozygote male containing the GDe and the SGyA transgene. No significant deviations between both the F2 and F3 data sets were found (p value = 0.3446). g Inheritance of the GDe in F2 and F3 progeny from crosses involving a transheterozygote male containing the GDe and the autosomal Cas9 transgene. Significant differences between GDe inheritance observed in F2 progeny (when compared to control; p value = 0.0129) and GDe inheritance between F2 and F3 progeny (p value = 0.0051) Blue shaded boxes in crossing schematics highlight instances where a bias of GDe transmission is observed. Green arrows indicate the conversion of WT PolG2 allele into the GDe. Gray numbers represent the expected Mendelian inheritance percentages of the GDe. Green percentages indicate the homing/bias of GDe. At least 18 experimental/replicate crosses were set up per Cas9 experiment, and 6 for the control to determine F2 progeny outcomes. For F3 outcomes, at least nine experimental crosses were performed. Significance was determined using a two-tailed unpaired student’s t test. For inheritance and male bias plots, vertical bars represent SEM. n.s. represents Non-significant. ***p < 0.0005; **p < 0.001; *p < 0.01. Source data is provided as a Source Data file.

In assessing the ability of the SGyA element to promote the non-Mendelian transmission of the GDe element, we observed that when F1 SGyA/GDe transheterozygous ♂’s were outcrossed to WT ♀’s, 65.3% of the F2 offspring, on average, inherited the GDe element (marked by the dominant GFP marker) compared to 50.6% of the F2 progeny in the negative control (F1 heterozygous GDe ♂’s outcrossed to WT ♀’s; p value = 0.0042, unpaired t test) (Fig. 4e). Similarly, when transheterozygous ♂’s (with an autosomal source of vasa-Cas9 and GDe) were outcrossed to WT ♀’s to assess GDe inheritance, we found that 69% of the F2 offspring on average inherited the GDe element. (Fig. 4e). This result is similar to the previous study which found that 63% of the offspring inherited the GDe in crosses involving a WT mother and a transheterozygous father containing an autosomal source of vasa-Cas9 and the GDe²⁶. There were no significant differences in inheritance rates of the GDe transgene between SGyA and autosomal Cas9 experiments (p value = 0.337, unpaired t test) (Fig. 4e). A subsequent outcross was carried out with the F2 transheterozygous progeny to determine if the homing rate of the GDe changed in subsequent generations (in the F3 progeny). We did not observe a significant deviation of GDe inheritance frequencies between the F2 and F3 progeny data of SGyA and autosomal Cas9 (Respectively, Fig. 4f, g). Taken together, these data suggest SGyA can function as a split gene drive and has comparable drive efficiency in the male germline to an autosomal source of Cas9.

Modeling indicates SGyA-based drive systems enact enduring population modification and rapid suppression

Advancing upon the characterization of Y-encoded Cas9 functioning as a split gene drive, and the goal of utilizing Y-encoded Cas9 as a population suppression system, we performed modeling to explore the potential for SGyA-based drive systems to enact efficient population modification and suppression. We conducted population simulations using the MGDrivE framework³⁵, comparing the performance of the Y-linked systems to equivalent X-linked and autosomal systems (Fig. 5). We performed simulations for Anopheles gambiae, a mosquito disease vector that proof-of-concept gene-editing tools from D. melanogaster are often applied to^2,3. Two populations with an equilibrium size of 10,000 were simulated, exchanging migrants at a rate of 1% per mosquito per generation³⁶. For all drive systems, 12 consecutive weekly releases of 10,000 ♂ mosquitoes homozygous or hemizygous for each drive allele were simulated in the release population, and spread in both the release and neighboring populations was recorded.

a Model predictions for releases of An. gambiae mosquitoes homozygous or hemizygous for three different split drive systems intended for population modification. The SGyA-based system in which the Cas9 allele is Y-linked (left) is compared with an autosomal split drive system (middle) and a system in which the Cas9 allele is X-linked (right). In all cases, the gRNA/effector allele is autosomal. Life-history and gene drive parameters are provided in Supplementary Data File 4. 12 weekly releases were simulated in a population with an equilibrium size of 10,000 adults and a 1% per mosquito per generation migration rate with a neighboring population of the same equilibrium size. Model predictions were computed using 100 stochastic realizations of the MGDrivE framework³⁵. Total adult female population size (dark blue), adult females carrying at least one copy of the gRNA/effector allele (red), adult females without the gRNA/effector allele (purple), and adult females carrying at least one copy of the Cas9 allele (light blue) were plotted for each system. Notably, for the Y-linked split drive system, the gRNA/effector allele persists longer in the population than for the autosomal or X-linked split drive systems. The Y-linked system also spreads to a higher frequency in the neighboring population. b Model predictions for equivalent releases of three population suppression systems: an SGyA-based Y-linked X-shredder (left), an autosomal homing-based drive targeting a gene required for female fertility (middle), and an autosomal X-shredder (right). Simulations assumed high rates of DNA cleavage and low rates of resistant allele generation, as required for effective population suppression⁴⁰ (Supplementary Data File 4). Total adult female population size (dark blue), adult females carrying at least one copy of the intact drive allele (red), and adult females without the intact drive allele (purple) were plotted for each system. Both the Y-linked X-shredder and autosomal homing-based drive targeting a female fertility gene achieved population elimination in >97% of simulations. The autosomal X-shredder leads to transient population suppression at the release site and limited spread to the neighboring population.

We first compared the performance of an SGyA-based split drive, in which the Cas9 is Y-linked and the gRNA locus is autosomal, to split drive systems in which the Cas9 is: (i) at an unlinked autosomal locus, and (ii) X-linked (Fig. 5a). For standard An. gambiae life-history parameters and gene drive parameters from a split drive system engineered in another mosquito vector, Ae. aegypti³⁷ (confinable split drive systems have yet to be demonstrated in An. gambiae) (Supplementary Data File 4), modeling results suggest that 12 weekly releases are sufficient to drive the gRNA/effector allele (red) to high frequency in the release population (>95% of ♀’s having at least one copy of the effector gene) for all three split drive designs. The Cas9 allele (blue) then falls out of the population due to a fitness cost, and the gRNA/effector allele is slowly eliminated as it also has a fitness cost and its inheritance bias is dependent upon co-occurrence of the Cas9 and gRNA alleles. Two interesting distinctions between the Y-linked and autosomal/X-linked split drive systems are that: (i) the gRNA/effector allele persists in the population for longer for the Y-linked system, and (ii) the Y-linked system spreads to a higher frequency in the neighboring population. An important metric for population modification strategies is the “window of protection” (WoP), which we define here as the duration that ♀ mosquitoes having the anti-pathogen effector gene remain at a frequency of 90% or higher in the population. When Cas9 is Y-linked, we calculate a WoP of 672 days from our simulations, which is significantly higher than the WoP for the autosomal split drive (428 days) and when Cas9 is X-linked (344 days). Migrants having the gRNA/effector allele accumulate in the neighboring population. Consequently, for the Y-linked design, the gRNA/effector allele spreads to a higher frequency and persists for a longer duration in the neighboring population, reaching a maximum carrier frequency (frequency of ♀ mosquitoes having at least one copy of the allele) of 48%, compared to 34% for the autosomal design and 25% for the X-linked design. The Y-linked design is, therefore, less confineable, although all three designs are self-limiting, meaning that spread in both the release and neighboring populations is transient.

Next, we compared the performance of an SGyA-based X-shredder, in which all drive components are Y-linked, to two other population suppression drive systems: (i) an autosomal X-shredder, and (ii) an autosomal homing-based drive targeting a gene required for ♀ fertility⁸ (Fig. 5b). The Y-linked X-shredder is a promising population suppression system for An. gambiae mosquitoes (for which ♂ are XY) as cutting of X gametes in the ♂ germline leads to an increasingly ♂ sex bias and potentially a population crash or persistent population suppression^15,38. As this system spreads from a low population frequency, it could be effective over a wide geographic range. A general weakness of population suppression strategies is that drive-resistant alleles have a significant selective advantage, and hence if they emerge, the population is likely to rebound³⁹. Assuming high rates of DNA cleavage (0.99 per heterozygote), as seen for other CRISPR-based drive systems in An. gambiae⁷, and a low rate of resistant allele generation (10⁻⁶ per heterozygote), as required for effective population suppression⁴⁰ (Supplementary Data File 4), we find that 12 weekly releases lead to population elimination in 99% of the simulations, which is reached on average 21 weeks after the final release. A notable difference for the autosomal homing-based drive targeting a ♀ fertility gene is that, for equivalent parameters (Supplementary Data File 4), it achieves population elimination with similar frequency (98% of simulations), but reaches elimination more slowly, on average 36 weeks after the final release. That said; for lower resistant allele generation rates (<10⁻⁶ per heterozygote), both systems are expected to achieve population elimination ~100% of the time. Additionally, the Y-linked X-shredder, by targeting multiple genetic loci on the X chromosome simultaneously¹⁰, may be easier to limit resistant allele generation for. Lastly, the autosomal X-shredder is a self-limiting population suppression system that could be used as an alternative to SIT strategies, or as an intermediate technology prior to the release of a non-localized Y-linked X-shredder. Here, we see that 12 weekly releases of this autosomal system lead to transient population suppression at the release site (the population rebounds to 95% of its original size within 51 weeks) and only limited spread to the neighboring population.