A CRISPR endonuclease gene drive reveals two distinct mechanisms of inheritance bias

RNA guided CRISPR gene drives have shown the capability of biasing transgene inheritance in multiple species. Among these, homing endonuclease drives are the most developed. In this study, we report the functioning of sds3, bgcn, and nup50 expressed Cas9 in an Aedes aegypti homing split drive system targeting the white gene. We report their inheritance biasing capability, propensity for maternal deposition, and zygotic/somatic expression. Additionally, by making use of the tight linkage of white to the sex-determining locus, we were able to elucidate mechanisms of inheritance bias. We find inheritance bias through homing in double heterozygous males, but find that a previous report of the same drive occurred through meiotic drive. We propose that other previously reported ‘homing’ design gene drives may in fact bias their inheritance through other mechanisms with important implications for gene drive design.


Introduction
Genetic modification of wild populations has been proposed as a means of addressing some of the world's most pressing public health challenges and may be achieved through gene drive. Gene drive is the ability of a genetic element to bias its own inheritance, which allows it to spread a genetic change through a population without necessarily conferring a fitness benefit ('selfish DNA') 1 . There are many examples of this phenomenon in nature, acting through many different mechanisms 2 . Some types of gene drive rely on the action of sequence-specific DNA nucleases (enzymes that create DNA breaks). Double-stranded DNA breaks are a common occurrence in cells, and a range of mechanisms exists to repair the DNA damage. Correspondingly, different nuclease-based gene drives can potentially bias their inheritance through different mechanisms. The development of synthetic nuclease gene drives received much attention 3,4 following the discovery of programmable nucleases in CRISPR systems 5 . Particularly the development of gene drives based on 'homing' (homing drives) and drives that cause the loss of non-drive bearing gametes or offspring (here referred to as meiotic drive).
Generally, in diploid organisms, one chromosome of each homologous pair is contributed by each parent, and each allele has a 50% chance of being passed along to a given progeny. Synthetic homing and meiotic endonuclease gene drives both rely on selectively creating double-strand DNA breaks on the non-drive-bearing homolog when the drive is present on one of the pair but not the other (heterozygous). Through different mechanisms, this results in an inheritance bias of an allele or genomic region and for meiotic drive potentially the whole chromosome. Meiotic endonuclease drives lower the inheritance of the competing chromosome within a pair by damaging it such that gametes carrying the non-drive chromosome are eliminated during gametogenesis or, in some cases, produce non-viable offspring. This includes the disruption of specific essential genes in toxin-antidote meiotic drives [6][7][8] , or through more structural damage such as with chromosome 'shredder' meiotic drives 9,10 . Natural sex-linked meiotic drive systems have been reported in Aedes and Culex mosquitoes 11,12 . Synthetic shredder endonuclease meiotic drives have generally sought to exploit naturally present large-scale, potentially repeating, sequence differences between chromosome pairs 9,10 . In contrast to meiotic drives, for homing drives sequence homology between the drive element and the target on the paired chromosome is essential. Homing drives bias their inheritance by creating a DNA break on the 'recipient' homologous chromosome corresponding to where the genetic material of the homing drive is located on the 'donor' homologous chromosome. If the coding sequence for the drive is then identified as missing from the cut chromosome, the DNA sequences of the drive and associated sequences can be copied over during repair of the DNA break ('homology-dependent repair').
For most reports of synthetic homing drives, the method of quantifying inheritance bias (phenotypic scoring of progeny carrying a drive linked genetic marker) cannot differentiate between the underlying inheritance biasing mechanism  . The large differences in 'design rules' that have emerged between synthetic meiotic and homing endonuclease systems may have contributed to the expectation that, for any given neutral target drive, there should be little overlap in the mechanism. This is supported by a (small) subset of publications that have used marked chromosomes [35][36][37] , especially pre-CRISPR [38][39][40][41] , which allowed the homing and meiotic drive mechanisms to be differentiated. These studies did not report observing meiotic drive.
However, we noted evidence for meiotic drive in male Aedes aegypti with a homing CRISPR gene drive design recently reported in Li et al. 36 . in a statistically significant increased inheritance of the drive from male drive parents. We reanalysed the results of Li et al. for nup50 males taking into account the sex linkage and found that the observed inheritance bias in double heterozygous males proceeded exclusively through meiotic drive.
We set out to test the hypothesis that the apparent meiotic drive observed with the nup50 expression pattern is a more general phenomenon and also occurs with other A. aegypti gene drives that show activity in males. In collaboration with the original authors, we repeated the nup50-Cas9 crosses and performed crosses with Cas9 expression under the control of putative transcriptional regulatory regions from two additional A. aegypti germline genes. The first, suppressor of defective silencing 3 (sds3) has been shown, by dsRNA-induced knockdown in Anopheles gambiae, to be required for normal development of the ovarian follicles and testis, with no other obvious defects 42 . The second, benign gonial cell neoplasm protein (bgcn) is involved in regulating and promoting gametogenesis in both sexes 43 and has been described in the context of gene drive in Drosophila melanogaster with I-SceI 38 and A. aegypti 44 .
For each Cas9 expressing line, we report the degree of inheritance bias of the w GDe element for both sexes and, in males, the mechanism of inheritance bias. In addition, by scoring somatic eye phenotypes, we also find strong evidence of zygotic/somatic expression, maternal deposition and unexpectedly a currently unexplained effect of the Cas9 carrying grandparent's sex on w GDe grand-offspring phenotypes.

DNA constructs
The sds3-Cas9 construct was produced by making several alterations to plasmids provided by Omar Akbari 45 . The sds3 construct contains, within piggyBac terminal sequences, an insect codon optimised Cas9 followed by a T2A self-cleaving peptide and EGFP. To improve visibility of the fluorescent marker, the initial OpIE2-DsRED cassette was replaced with PUb-mCherry-SV40. To reproduce the germline-specific expression pattern of sds3, the Cas9:EGFP coding sequence is preceded and followed by the non-coding sequences flanking the endogenous sds3 gene's open reading frame. The 3' UTR is followed by an additional P10 3'UTR. To determine the 5' and 3' UTR of the sds3 gene 5' and 3' Rapid amplification of cDNA ends (RACE) was performed using the SMARTer R RACE 5'/3' Kit (Takara Bio) on RNA isolated using Trizol (Life Technologies) from female and male, 5-7 day post eclosion WT adult mosquitoes. RACE PCRs were performed using gene-specific primers: 5'-TGTGCTGTTCGTATGGTTCCGGATGG-3' then nested primer 5'-TCGTCCAGCAAAAGA ACCAACTGCCCAG-3' for 5' RACE and for 3'RACE 5'-ACGTCGACCTAATGAACCGCTTCCG-3' and nested primer 5'-GGGCAGTTGGTTCTTTTGCTGGACG-3', amplicons were cloned using the CloneJET PCR Cloning Kit (Thermo Scientific) and Sanger sequenced. In total, 1959 bp upstream of the translational start was amplified in order to include the most significant promoter elements (including 202 bp of 5'UTR). The 5' and 3' sequences were amplified from WT adult gDNA extracted using the NucleoSpin Tissue Kit (Machery-Nagel) by PCR using Phusion High Fidelity PCR Master Mix (NEB) and primers 5'-ttttgcggccgcTCTGTTTGAATATGTTTCCGAGAA -3' and 5'-ttttctcgagTTTCCGCGACAAAAACACAGA-3' which add 3/29 restriction enzyme recognition sites NotI and XhoI, respectively (underlined).The promoter fragment digested with NotI/XhoI, Cas9 digested with XhoI/FseI, and pBac PUb-mCherry-SV40 digested with NotI and FseI were ligated with T4 DNA ligase (NEB) in a three-way ligation. The 307 bp 3'UTR was then amplified from the same WT adult genomic DNA by PCR using Phusion High Fidelity Master Mix (NEB) and primers: 5'-ttttttaattaaGGAAACAAGGATCTCAACTCTCGAGC-3' and 5'ttttg cgatcgccctcgagcTTCTTAGGTACAATTGTAAAACATAGTT-3' to amplify from the stop-codon until just beyond the transcript end as determined by 3' RACE. This amplicon was digested and ligated into the sds3-Cas9 intermediate plasmid digested with PacI. See GenBank depository for the sequences and annotations of sds3. The bgcn construct was created similarly to the sds3 construct and is described in Anderson et al. 44 . The sequence and insertion site of the gRNA element (3xP3-tdTomato) and nup50 lines are described in Li et al. 36 . The bgcn and sds3 constructs use a Cas9 that is insect codon optimised as described in Anderson et al. 44 . The nup50 line makes use of a human codon optimised Cas9 45 .

Mosquito lines
A. aegypti Liverpool strain (WT) was a gift from Jarek Krzywinski. The nup50-Cas9 and white gRNA expressing element (w U6b-GDe , hereafter w GDe ) lines provided by Omar Akbari are described in Li et al. 36 . At Pirbright, the nup50-Cas9 line was maintained as a mix of homozygotes and heterozygotes with periodic selective elimination of wildtypes; the w GDe element line was reared as homozygous in our facilities. Cas9 expressing lines generated at the Pirbright facilities were maintained as heterozygotes, usually by crossing transgenic males to WT females.
All mosquito lines were reared in an insectary facility under constant conditions of 28 • C, 65-75% relative humidity and 12:12 light/dark cycle (1h dawn/1h dusk). Larvae were fed ground TetraMin flake fish food (TetraMin) while adults were provided with 10% sucrose solution ad libitum. Defibrinated horse blood (TCS Bioscience) was provided using a Hemotek membrane feeding system (Hemotek Ltd) covered with Parafilm (Bemis).

Crosses for homing assessment
Male and female adults, homozygous for w GDe were crossed with heterozygous mosquitoes of the Cas9 lines. Their progeny were screened as late larvae under fluorescence using an MZ165FC microscope. Eye phenotype was also evaluated. Double heterozygous mosquitoes carrying both transgenes were then crossed to WT mosquitoes. Inheritance of the transgenes as well 4/29 as eye phenotype, was again assessed under a fluorescence microscope. For the nup50-Cas9, double heterozygous females from each cross were allowed to lay eggs individually. For bgcn-Cas9 and sds3-Cas9 at least 20 double heterozygotes were crossed separately by sex such that double heterozygotes were always crossed to WT of the opposite sex. The exact number and phenotype of the progeny of each cross are shown in Table S3-S4. The individual cross data for nup50-Cas9 are shown in Table S7-S10. In some cases, F 1 double heterozygotes produced from the same cross presented with a different fluorescent marker or eye pigment phenotypes. In each case, these were noted in the cross tables, and examples of the phenotypes are shown in Fig S1.

Statistical analysis of w GDe inheritance bias
For each F 1 sex, the w GDe inheritance rate in the absence of a Cas9 expressing element (Table S6) was used as the baseline inheritance. This was 52% (620/1203) for males and 51% (308/605) for females. These rates were used as the expected outcome in a Fisher's exact test with the w GDe inheritance from F 1 parents that carried the w GDe and one of the Cas9 expressing elements. A significant difference in w GDe inheritance is taken as evidence for drive activity. See Table S11.

Statistical analysis of somatic expression and parental deposition
For each Cas9 line, the fraction of mosaic eyed (ME) or white-eyed (WE) progeny among the F 2 offspring inheriting w GDe but not the Cas9 ('+w GDe ;−Cas9') from F 1 drive males served as a control for the frequency of such phenotypes in the absence of somatic expression or maternal deposition. For somatic expression, the ME/WE fraction of the F 2 progeny harbouring both the Cas9 and w GDe elements from F 1 drive males was compared to the control cross using Fisher's exact test (Table S12). For maternal deposition, the F 2 progeny harbouring only the w GDe element from F 1 drive females was compared to the control (Table S13).

Statistical analysis of the influence of factors on the fraction of mosaic and white-eyed progeny
A generalised linear model with binomial errors was created that included Cas9 promoter (sds3, bgcn, nup50), F 2 Cas9 status (+/−), F 2 sex (♂/♀), F 1 drive parent sex (♂/♀), and F 0 Cas9 carrying grandparent sex (♂/♀). The response variable was the proportion of ME and WE progeny among the all the F 2 progeny from that cross and F 2 sex (48 conditions). The analysis was performed in R version 4.0.2 using the glm function. See Table S14.

Statistical analysis of homing and meiotic drive
For homing, the background recombination rate (calculated from the male F 1 +w GDe ;−Cas9 cross Table S6) is used as the expected outcome in a Fisher's exact test. For the control cross (in the absence of possible Cas9 mediated inheritance bias) the w GDe allele was provided by the male F 0 grandparent and therefore M-linked in the male F 1 s. In the absence of recombination, all male F 2 s should be w GDe positive, and all female F 2 s should be w GDe negative. Out of the 1203 progeny scored, we saw 13 (1.08%) recombination events. 2 out of 609 male F 2 s were w GDe negative, and 11 out of 581 female F 2 s were w GDe positive.
For the crosses including a Cas9 element, a significant increase in recombination rate between the recipient/donor chromosome 5/29 marker (sex) and the drive element was taken as evidence of homing (Table S15). For meiotic drive, a statistically significant difference in the inheritance of either the recipient or donor chromosome (i.e. F 2 sex) is taken as evidence for meiotic drive (Table S16). The progeny sex ratio is compared to the sex-ratio in the absence of a Cas9 expressing element (Table S6).  For male F 1 parents, the w GDe allele should be found exclusively in only one sex of their F 2 progeny, apart from background recombination events, or drive induced recombination (homing).

Results
To assess the degree and, in males, the mechanism of inheritance bias, we bred transgenic A. aegypti mosquitoes to create and analyse a 'split drive' arrangement. In this split drive, the w GDe allele expresses a gRNA targeting the wildtype white gene (w + ) at the site corresponding to where the drive element has been inserted into and disrupts its protein-coding sequence (Fig   1a). To create individuals in which drive can occur, the w GDe element is combined with the other component of the split drive,  Figure 2. Gene drive element (w GDe ) inheritance and somatic eye phenotype in the progeny of double heterozygote split-drive carriers ('Drive parents'). a Parental germline inheritance bias of w GDe when combined with a nup50, bgcn, or sds3-Cas9 expressing element. For each of the three Cas9 regulatory elements, the inheritance rates are reported in columns left-to-right: Male drive F 1 s with paternal Cas9 contribution, male drive F 1 s with maternal Cas9 contribution, female drive F 1 s with paternal Cas9 contribution, and female drive F 1 s with maternal Cas9 contribution. The horizontal dotted line indicates the expected Mendelian 50% inheritance from heterozygous carriers. For nup50, individual crosses were performed, and each circle represents the percentage of w GDe positive progeny from an individual parent. Error bars are the Wilson confidence intervals for the binomial proportion calculated for the pooled progeny count, which does not take into account the potential lack of independence due to individual parent 'batch' effects. Stars indicate statistical significance as presented in Table S11. b The percentage of +w GDe progeny that display a mosaic or total loss of eye pigment phenotype. F 2 progeny are segregated by the drive carrying F 1 's sex (♂, ♀), the F 2 's Cas9 transgene inheritance (−Cas9, +Cas9), the Cas9 regulatory sequences (sds3, bgcn, nup50), and the F 2 's sex (♂, ♀). The circle size indicates the number of progeny that make up that group, and circle colour indicates if the Cas9 carrying F 0 grandparent was male (Blue) or female (Orange). The set of progeny that came from F 1 drive females are indicated with 'Maternal Deposition'. The set of progeny that inherited both a w GDe allele and Cas9 element are indicated with 'Somatic Expression'. Within matched crosses (each row), differences in the white phenotype rate corresponding to the Cas9 carrying F 0 's sex are referred to as a grandparent enhanced somatic phenotype. White phenotype rates for −w GDe progeny are shown in Fig S2.

8/29
In Table S2-S4, the progeny genotype and phenotype of the F 2 progeny are reported for each cross. In Table S11 and Fig 2a the percentage of w GDe inheriting F 2 progeny from double heterozygote F 1 parents are reported for each Cas9 expressing line. The inheritance rates are split by the sex of the double heterozygous parents, and whether those parents inherited the Cas9 element from the F 0 grandmother (maternal) or F 0 grandfather (paternal). For each condition, Fisher's Exact tests were performed comparing the w GDe inheritance rates to those in the absence of any Cas9 element for male (52%, 620/1203) or female (51%, 308/605) parents (Table S6) For nup50-Cas9, the progeny were collected individually from F 1 parents (Table S7-S10). As can be seen in Fig 2a, there was considerable variation between the inheritance rate from different parents carrying the same drive, a notable feature that was reported in many other drive papers 18,24,26,35 . Due to this over-dispersal, we cannot reliably determine if there is a statistical difference in the inheritance rate between the different Cas9 regulatory elements. However, because this over-dispersal is expected only to occur if the drive is functional, our method for determining a difference from the control remains valid, albeit with a potentially inflated false-negative rate.
All progeny were evaluated for eye pigment defects which may result from embryonic or later somatic bi-allelic disruption of the white gene by the w GDe element and NHEJ mutations. Since the double heterozygote drive carrying parents were crossed to wildtype individuals, each progeny inherited at least one dominant functional white allele from the non-drive parent, and, if the w GDe element is not inherited, potentially an additional one from the drive parent. Bi-allelic loss of function of the white gene must therefore occur through deposition into, or somatic expression in, F 2 individuals. The progeny from the −Cas9 control crosses did not present with a white phenotype (Table S6). The eye pigment phenotype for the three Cas9 expressing lines is reported in Fig 2b ( For the male double heterozygotes sds3-Cas9 crosses, of the F 2 progeny (♂ and ♀ pooled) that inherited both the w GDe and Cas9 element 86% of presented with a mutant somatic phenotype if the Cas9 carrying F 0 was male, and 98% if the Cas9 carrying F 0 grandparent was female (F 1 :♂, +Cas9 in Fig 2b and Table S12). For bgcn-Cas9 this was 7/17%, and for nup50-Cas9 this was 95/98%. However, if only the w GDe element was inherited, no cross had more than 1% of the pooled ♂ and ♀ F 2 progeny present with a somatic phenotype, presumably resulting from the lack of paternal Cas9 transmission through the sperm (F 1 :♂, −Cas9 in Fig 2b and Table S12). For each cross, this was a significant difference (Table S12) indicating 9/29 somatic expression, without substantial paternal deposition of Cas9/Cas9:gRNA w . In contrast to the <1% rate observed in the progeny of F 1 drive males, the crosses with female double heterozygotes where only the w GDe element was inherited, 40/95% (F 0 :♂/♀) of the bgcn, and an astounding 99/100% of the sds3 and 100/99% of the nup50 progeny presented with visible somatic phenotypes (F 1 :♀, −Cas9 in Fig 2b and Table S13). This indicates strong maternal deposition of Cas9/Cas9:gRNA w . For each cross, this was a significant difference (Table S13). Maternal Cas9 deposition without substantial paternal deposition has been reported for many other drive systems 13-15, 18, 24-26, 29, 30, 36, 50-53 .
Surprisingly, in the w GDe inheriting progeny we observed a trend where a higher fraction of progeny exhibited a somatic phenotype when the Cas9 carrying grandparent was female as opposed to male (F 0 :♂vs F 0 :♀in Fig 2b). Contrasting each male F 0 Cas9 carrying grandparent cross with the equivalent cross with a female F 0 Cas9 (each row in Fig 2b) showed, for female F 0 Cas9, an average 5.2% (sd:14.4%) percentage point increase in white/mosaic eyed phenotype among +w GDe F 2 progeny. While maternal deposition from a Cas9 carrying grandparent may increase the number of w GDe and NHEJ mutated alleles passed along by the F 1 parental generation to −w GDe progeny (S2), this should not, in contrast to what we observe, influence the phenotype of the progeny that inherit the w GDe element (2b). If the w GDe element is inherited there is no opportunity to inherit a germline NHEJ mutation that was created due to deposition from the grandparent into the parent. We created a generalised linear model that included Cas9 promoter, F 2 Cas9 status, F 2 sex, F 1 drive parent sex, and F 0 Cas9 carrying grandparent sex (Table S14). The sex of F 0 Cas9 carrying parent had a significant influence on the fraction of white/mosaic eyed +w GDe F 2 progeny. We termed this phenomenon Grandparent Enhanced Somatic Phenotype (GESP). All other factors were significant too, apart from the sex of the F 2 progeny.
In A. aegypti, the white gene is tightly linked to the sex-determining locus. This locus comprises two forms, a dominant male-determining allele M and a corresponding m, such that males are M/m and females m/m. While the molecular basis of sex determination in this mosquito is not fully understood, M is associated with Nix, a gene shown to be involved in sex determination 54 . Analogous to an XY chromosome system, male offspring of an M/m male always carry the paternal M allele and female offspring the paternal m, with no such distinction between the two m alleles of the mother. For the male parent, if the initial linkage of w GDe to either m or M is known (determined by the sex of the w GDe carrying grandparent), the sex of the progeny can be used as an indication of whether an observed inheritance bias is due to new recombination events (homing), or increased inheritance of the original drive carrying chromosome (meiotic drive) (Fig 3a). To this end, we stratified the w GDe inheritance by the sex of the F 2 progeny for each of the male double heterozygous parents (Fig 3b).
The background recombination rate of w GDe and sex in the absence of any Cas9 element was 1.08% (13/1203) (Table S6) and was compared by Fisher's Exact tests to the recombination rate from w GDe Cas9 male double heterozygotes (Table S15).
As reported above, only one cross each of the sds3 and bgcn double heterozygotes showed a significant increase in overall w GDe inheritance. However, quantifying conversion with marked chromosomes is much more sensitive than measuring overall w GDe inheritance rate.
For the sds3 double heterozygous males with paternal Cas9 contribution, 9% of their progeny were w GDe positive males.   Figure 3. Separating w GDe inheritance by F 2 sex allows different mechanisms of inheritance bias to be distinguished. a Illustration of how homing, meiotic drive and copy-grafting/co-conversion are expected to influence the observed sex-linkage of an M linked w GDe element in the progeny of male drive double heterozygous parents. b Parental germline inheritance bias of w GDe when combined with no Cas9, nup50, bgcn, or an sds3-Cas9 expressing element. We included the nup50 results from Li et al. that use the identical nup50-Cas9 line. For each of the three Cas9 regulatory elements, the w GDe inheritance from male double heterozygotes is reported in pairs of columns segregated by the sex of the F 2 progeny. In each case, the first pair of columns are the results for when w GDe is m-linked, and the second pair are the results for when w GDe is M-linked. Error bars are the Wilson confidence intervals for the binomial proportion calculated for the pooled progeny count. The overlaid numbers are the percentage (cumulative within each column) of the indicated F 2 sex and w GDe status among all progeny from that cross.

11/29
This indicates that 19% (41/216 p-value: <0.001 *** ) of the recipient chromosomes were converted by the combined effect of homing and background recombination. The same was true for maternally contributed Cas9 where 9% of w GDe progeny were female, indicating 19% (19/102 p-value: <0.001 *** ) of recipient chromosomes were converted. For bgcn males with paternal Cas9, we found that 3% of their progeny were w GDe positive males which indicates 5% (24/464 p-value: <0.001 *** ) of the recipient chromosomes were converted. For bgcn with maternally contributed Cas9, 8% of female progeny carried w GDe , indicating a conversion rate of 20% (32/160 p-value: <0.001 *** ). This large difference in homing rate between crosses with maternal vs paternal F 0 Cas9 suggests that for bgcn maternally deposited Cas9 may contribute more to homing than autonomously expressed Cas9. Low expression, but high maternal deposition rate of bgcn is also consistent with the difference in the observed white phenotype rate (2b). However, the difference in homing events accounts for less than half of the overall difference in w GDe inheritance rate between maternally (66%) and paternally (50%) contributed bgcn Cas9. This may suggest that another inheritance biasing mechanism is active.
For the nup50 double heterozygote males with paternal Cas9 contribution, 12% of their progeny were w GDe positive males, indicating 24% (210/869 p-value: <0.001 *** ) of the recipient chromosomes were converted by homing. For maternally contributed Cas9 this was 11% of progeny and 23% (315/1387 p-value: <0.001 *** ) of recipient chromosomes. We also performed this analysis on the nup50 crosses reported in Li et al. (Table S5) (Table S16). This sex-bias should not occur through homing. Instead, this seems consistent with a meiotic drive mechanism where some of the non-w GDe chromosomes are lost, or conversion of a very large region encompassing both w GDe and the sex-determining region (Fig 3a). For the crosses performed for this study, including the nup50 line, no significant difference in sex, and by extension recipient vs donor chromosome inheritance, was detected (Table S16). For bgcn with maternal F 0 Cas9, 59% of all F 2 s were male, but this did not rise to our significance threshold due to the relatively low number of progeny scored for this cross.

Discussion
In this study, we report the efficiency and mechanisms of three CRISPR-Cas9 nuclease gene drives targeting the white gene, expanding the tool-set for developing genetic control strategies for the public-health relevant Aedes aegypti mosquito. In our hands, sds3, bgcn, and nup50 expressed Cas9 each resulted in increased inheritance of the w GDe drive element, with the primary mechanism seeming to be homing. In addition, for each promoter, we find evidence of maternal deposition and somatic expression and, unexpectedly, a currently unexplained effect of the Cas9 carrying grandparent's sex on the grand-offspring phenotypes (Fig 2b)  white locus to be a good drive target, allowing for efficient transmission bias and convenient readout of an easily-scored visible recessive phenotype 36 . In addition, the locus allows for effective transgene expression from a sex-linked locus which may be of particular use for future drives and other genetic control approaches. For the bgcn drive in males, the recipient chromosome conversion rate was much higher with maternally contributed Cas9 (19%) compared to paternally contributed Cas9 (5%). These results suggest that, in at least males, the bgcn drive may function primarily through maternally contributed Cas9. Homing through Cas9 deposition in the absence of expressed Cas9 ('shadow drive') has been reported for other drives 29,30,35 , but to our knowledge not as the primary means of inheritance bias for a drive. We find nup50 and sds3-Cas9 capable of directing transmission bias in females and males, and we did not find that maternal deposition from the Cas9 carrying grandmother negatively influenced the homing rate observed in males.
For all drives, the almost complete absence of any somatic phenotype in individuals that did not inherit the w GDe element ( Fig S2) could indicate that, while maternal deposition of the Cas9 occurs, the gRNA w or gRNA w :Cas9 complex are either not deposited or rapidly degraded. However, progeny that did not inherit the w GDe element instead inherited the (initially) w + allele from the double heterozygous parent. For mosaic eye phenotypes to occur in these individuals, up to two functional w + alleles may need to be disrupted by deposition instead of one; direct comparison of the rates of somatic mutation between offspring that do and do not inherit the gRNA w transgene are therefore potentially misleading. Moreover, some non-w GDe progeny may have inherited a white allele that contained a functional, but cut-resistant, NHEJ mutation (type-1 resistant mutation) which would make bi-allelic disruption impossible. For the −w GDe F 2 progeny, maternal deposition from the F 0 grandmother could increase their probability of inheriting a mutated w allele from their F 1 parent. As such, GESP does not apply, and only refers to +w GDe F 2 progeny where the sex of the w GDe or Cas9 carrying grandparent influences their propensity to present with a somatic phenotype. While deposition from a F 0 grandparent may explain a change in the quantity of w GDe alleles passed along by the F 1 drive parent (shadow-homing), it does not seem to explain a change in the phenotype of those F 2 progeny that inherited a drive element (GESP). Genomic imprinting or transgenerational persistence of the deposited Cas9 mRNA/protein may underlie GESP.
For nup50 the overall inheritance biasing rate and somatic/embryonic drive activity (≈100%) closely match those reported by Li et al. 36 and underscore its potential utility for systems such as precision-guided SIT 55 . Yet, an important finding of our work is the propensity of this drive to function through two different mechanisms. The selective inheritance or elimination of a chromosome is generally achieved through creating multiple DNA breaks on the target chromosome 10, 56-58 (e.g. X-shredder) or by disrupting an essential gene 7,8 . Meiotic drive through a single cut in a non-essential gene as found by Li et al., and reported here, is noteworthy. One explanation could be the chromosomal location of the induced double-stranded break. A single cut has been demonstrated to be sufficient for inheritance bias through the loss of a chromosome in yeast when it is targeted to a centromere, while nearby sites were not sufficient 59 . The white gene is located relatively near the centromere. However, a centromere effect does not explain the difference in results from this study and that of Li et al., which instead suggests subtle differences in the rearing conditions or background genetics of the mosquito strains may have a significant influence on the 13/29 underlying mechanisms. Gene drive assessment performed in D. melanogaster with different genetic background revealed differences in their activity but changes in the underlying mechanism were not investigated 25 . The nup50-Cas9 and w GDe transgenic lines used in this study are the same as described in Li et al., but the crosses to assess homing were made to LVP strains maintained for a long period of time in different insectaries. Mosquito colonies maintained in laboratories can suffer from founder and drift effects, affecting their genetic background and reducing their heterozygosity 60 . Moreover, genetic variability in A. aegypti colonies from the same strain but reared in different laboratories has been documented 61 .
We cannot rule out that the sex-bias we report for Li et al.'s nup50-Cas9 is due to copying of a >45 Mbp 48, 49 region comprising both the w GDe and the sex-determining region (Fig 3a). While co-conversion of sequences not directly within a drive element has been reported 35,41 , the large distance between the w GDe drive and M locus leads us to believe this is less plausible. Moreover, similar homing drives have been reported to be sensitive to the alignments of the homology arms 8, 20, 24 , and several studies have reported partial homing events [22][23][24] . These partial homing events are seemingly due to sequences in the drive element (such as the gRNA gene) having undesired homology to the recipient chromosome and results in only part of the drive element being copied over. Similarly, a toxin-antidote CRISPR gene drive element was not copied despite targeting a nearby site on the homologous chromosome 8 . These results are inconsistent with a single DNA-break inducing large scale homing beyond the (immediately) adjacent regions of homology.
To our knowledge, for drive designed to function through homing recipient/donor chromosome markers have been used with non-CRISPR nucleases in D. melanogaster [38][39][40] and An. gambiae 41 and with CRISPR-Cas9 in D. melanogaster 35 , A.
aegypti 36 and Mus musculus 37 . Collectively, the publications with D. melanogaster provide significant evidence against meiotic drive mechanisms contributing significantly to the observed inheritance bias of the 'homing' drives examined. Moreover, nos-Cas9 has been reported to cause inheritance bias in many gene drives with a homing drive design 20, 22, 24-26, 29, 30, 32 but failed to do so in a meiotic drive design: despite having far higher cleavage activity, a nos-Cas9 X-shredder meiotic drive did not result in sex-bias, while β tub85D expressed Cas9 did 58 . There may however be an exception, a study of drives with a homing design noted a reduced transmission of the recipient chromosome for a set of crosses where the split Cas9 transgene happened to function as a chromosome marker 52 . It should be noted that the drives in this study targeted essential genes, potentially complicating the interpretation of the mechanism of inheritance bias.
In light of our results, re-evaluation of the A. gambiae I-SceI gene drive reported by Windbichler et al. may suggest that a meiotic drive effect in homing drive designs is more widespread 41 . Their drive carrying line had a small marker (NotI restriction site) located approximately 0.7 kilobases from the I-SceI cut-site on the recipient chromosome, but not on the donor drive chromosome. They reported 86% inheritance of the drive element from heterozygote males. However, drive alleles that included the NotI site only accounted for around half the increased drive allele inheritance. The authors attributed this discrepancy to co-conversion, where homing of the drive element also replaced the nearby NotI marker. A combined meiotic drive and homing effect would seem to provide an alternative explanation. In the M. musculus drive reported by Grunwald et al.
the recipient chromosome had a linked coat colour marker that allowed homing events to be precisely tracked 37 . In females,

14/29
vasa-Cre induced CAG-Cas9 expression and resulted in homing rates of 42% (36/86) and 11% (5/47) depending on the Cas9 insertion site. In males, no homing was observed with any drive. However, for the vasa drives, males passed along the donor drive chromosome to 63% (45/71) and 54% (49/91) of their progeny, potentially indicating a meiotic drive mechanism in that sex. It should be noted that detecting meiotic drive using this method is less sensitive than detecting homing and more progeny would need to be scored to have confidence in this trend. Together, the A. aegypti 36 , An. gambiae 41 and M. musculus 37 drives indicate that meiotic drive in drives intended to function through homing may be a widespread occurrence. Distinguishing these mechanisms requires linked markers, and for some organisms, this type of study may be best reserved for drives that after initial tests warrant further development.
Our work further expands the Cas9 expression patterns that have been tested in the context of mosquito gene drives. It is notable that the drives with a homing design reported in the Anopheles mosquitoes A. gambiae [16][17][18][19] and A. stephensi [13][14][15] almost invariably have a dramatically higher conversion rate than those found in A. aegypti. It is not clear what underlies this difference. However, the fact that the modest conversion rate for nup50-Cas9 males remains stable despite a change in the mechanism may limit the possible explanations. This stability suggests that the factor(s) negatively affecting the conversion rate in A. aegypti are not specific to either homing or meiotic drive. Moreover, it also indicates that the difference in conversion rate observed between mosquito species is likely not due to the species favouring one mechanism over the other. Yet, the difference in mechanism between homing and meiotic drive through gamete destruction has important practical implications: First, the loss of gametes through a meiotic-drive mechanism may negatively affect mating competitiveness by lowering the number of viable gametes, though in some cases gametes may be produced in sufficient excess for this not to be significant. The homing mechanism functions through conversion, and should not affect gamete numbers. For the nup50 meiotic drive reported in Li et al., male nup50-Cas9 fecundity was tested and found to not differ from wildtype 36 . Second, on a 'per-cut', basis meiotic drive is moderately less efficient than homing. When meiotic drive removes a non-drive gamete/embryo, it thereby improves the chances of the remaining gametes/embryos. These may, in addition, to drive carrying gametes, include other wildtype and cut-resistant allele carrying gametes that were not destroyed. In contrast, homing converts a non-drive gamete to a drive gamete which does not benefit any of the leftover non-drive gametes making homing more efficient. Third, the linkage between different drive components may change very significantly depending on the mechanisms: for instance, if in a split-drive system the Cas9 is located near the gRNA element homing would still only increase the number of gRNA alleles, but not the Cas9 alleles. However, meiotic drive would increase the inheritance of both the gRNA and Cas9 element. This could theoretically cause a split-drive or daisy-chain drive 62 to spread more than anticipated. Locating each element on separate chromosomes would prevent this, and our data suggest that this may be a wise precaution to increase the predictability of their invasiveness.
Although, if anticipated or identified in early-stage field trails, a meiotic drive induced linkage between elements could also be leveraged, lowering the required release frequencies 63 . Nonetheless, in regards to risk-assessment of rare recombination events, the genomic distance at which two split-drive elements become strongly linked is presumably still much more permissive for a meiotic drive mechanism as opposed to a homing mechanism. Last, in the case of Li et al.'s white targeting A. aegypti

15/29
drive, its linkage to the sex-determining locus caused an otherwise neutral replacement drive to act, in males, like a sex-biasing suppression drive. This might be desirable for some applications, but obviously detrimental if the intended application were different. Most of these concerns apply even if the actual mechanism is co-conversion/copy-grafting of a large chromosome segment as opposed to meiotic drive.

27/29
−w GDe progeny with white-eyed or mosaic phenotype F 0 ♂ ♀  Figure S2. Somatic eye phenotype in the −w GDe progeny of double heterozygote split-drive carriers. The percentage of −w GDe progeny that display a mosaic or total loss of eye pigment phenotype. F 2 progeny are segregated by the drive carrying F 1 's sex (♂, ♀), the F 2 's Cas9 transgene inheritance (−Cas9, +Cas9), the Cas9 regulatory sequences (sds3, bgcn, nup50), and the F 2 's sex (♂, ♀). The circle size indicates the number of progeny that make up that group, and circle colour indicates if the Cas9 carrying F 0 grandparent was male (Blue) or female (Orange). The set of progeny that came from drive F 1 carrying females are indicated with 'Maternal Deposition'. The set of progeny that inherited a Cas9 element are indicated with 'Somatic Cas9 Expression'. Due to the sex-linkage of the w GDe element from F 1 drive males the number of F 2 progeny are unequally distributed among the groups for each sex. This sex-bias is more pronounced for the −w GDe progeny than for the +w GDe progeny due to the the effects of homing. Groups with <1 progeny are not shown. Within matched crosses (each row), differences in the white phenotype rate corresponding to the Cas9 carrying F 0 's sex cannot be solely attributed to a grandparent enhanced somatic phenotype for −w GDe progeny. White phenotype rates for +w GDe progeny are shown in Fig 2b. Note that inheritance of the w GDe element also prevents the potential inheritance from the drive parent of an undamaged wildtype white gene or a white gene mutation that retains its function (r1) in pigment production. As such, any difference in white phenotype rates between +w GDe and −w GDe progeny cannot be solely attributed to gRNA expression from w GDe in the F 2 progeny.