An. stephensi Indian strain (gift from M. Jacobs-Lorena, Johns Hopkins University) were maintained in insectary conditions (27 °C and 77% humidity) with a photoperiod of 12-h light/dark and 30 min of dawn/dusk. Sucrose solutions (10% wt/vol) were provided ad libitum and blood meals consisting of defibrinated calf blood (Colorado Serum Co., Denver) were offered to 3–7-day-old adults through the Hemotek® membrane feeding system. Larval stages were reared in distilled water and fed TetraMin® fish food mixed with yeast powder. Gene-drive mosquitoes were contained in ACL-2 insectary facilities at the University of California, Irvine and handled according to recommended safety procedures40,41,42.

Plasmids for the Swap strategy

The Swap strategy employed to convert nRec to Reckh uses the three plasmids shown in Fig. 1a: (1) pVG362_Aste-U6A-Swap3-gRNA to express the gRNA-sw3, (2) pVG363_Aste-U6A-Swap4-gRNA to express the gRNA-sw4, and (3) pVG344_Aste_kh2-MCRv3-vasa-Cas9 to provide the HDR template containing the recoded-kh coding fragment and the GFP marker.

To generate plasmids pVG362 and pVG363, a pair of oligonucleotides were synthesized (Integrated DNA Technologies) for each plasmid with 19 (pVG362) or 20 (pVG363) bases of the target sequence chosen for the strategy. These were annealed and ligated with T4 ligase (New England Biolabs) into the pVG145-Aste-U6A-Bbs1 plasmid8 linearized with BbsI. The cloning strategy was adapted from the work of Port et al.43. The oligonucleotides used to construct pVG362 were 1288_Aste-Swap3-Target_F (CTTGTTCTTGGAGGAGCGCACCA) and 1289_Aste-Swap3-Target_R (AAACTGGTGCGCTCCTCCAAGAA). The oligonucleotides used to construct pVG363 were 1290_Aste-Swap4-Target_F (CTTGTTACGttaattaaCGTAGAA) and 1291_Aste-Swap4-Target_R (AAACTTCTACGttaattaaCGTAA).

The pVG344 plasmid was cloned using the NEBuilder HIFI DNA Assembly Cloning Kit (New England Biolabs) to assemble four amplified fragments. Fragment 1 was generated by amplification of the backbone region of plasmid pVG163_pAsMCRkh28, fragment 2 was generated by amplification of the kh recoded rescue fragment from a plasmid synthesized by GenScript Inc., fragment 3 was amplified from a plasmid containing a 3× P3-GFP cassette commonly used for insect transgenesis, and fragment 4 also was amplified using pVG163_pAsMCRkh2 as a template. Primer pairs used to amplify each fragment were: Fragment 1: 494_pUC19_Backbone_F (GGTATCAGCTCACTCAAAGGCGGTAATACGG) and 1227_As-MCR2_GA_backbone_R (CGTAGAACGGAACCATCGCGTG), Fragment 2: 1231_As-MCR2_GA_RecodedFrag_F (CGCGATGGTTCCGTTCTACGG) and 1232_As-MCR2_GA_RecodedFrag_R (CTACGCCCC,,CAACTGAGAGAACTC), Fragment 3: 1230_As-MCR2_GA_GFP_F (TCTCTCAGTTGGGGGCGTAGCGTACGCGTATCGATAAGCTTTAAGATAC) and 1229_As-MCR2_GA_GFP_R (CACCGGTCGCCACCATGGTGAGCAAGGGCGAGGAGCTGTTCAC, Fragment 4: 1228_As-MCR_GA_backbone_F (CACCATGGTGGCGACCGGTGGATC) and 1241_As_MCR2_GA_HA2_R (CGCCTTTGAGTGAGCTGATACCGTGAGCAAAAGGAGACGG).

Microinjections and establishment of Reckh

Embryos were obtained from heterozygous females of the An. stephensi AsMCRkh2 (nRec) gene-drive line8. Microinjection procedures44 were performed using a plasmid mix containing 600 ng/μL of pRec-kh donor (pVG344) and 200 ng/μL of each gRNA-sw3 (pVG362) and gRNA-sw4 (pVG363) plasmids. Surviving G0 adults were sorted in pools of 2-4 males and 7-10 females and outcrossed to 10x WT females and 1× WT males, respectively. G1 progeny were screened as larvae for the inheritance of the GFP eye marker and kept as separate lines according to their male (♂4) or female (♀4) founder lineage. The two lines were screened routinely as larvae for the inheritance of the GFP eye marker and as pupae for the eye-color phenotype (black [WT], white, or mosaic) and maintained by intercrossing GFP+ black-eyed individuals. A homozygous drive line was established from the ♀4 intercrossed line.

Molecular confirmation of HDR-mediated target site integration of the Reckh cargo was performed on genomic DNA extracted from single GFP+ black-eyed individuals using the Wizard® genomic DNA purification kit (Promega). Primers Kh1-ext-fw (CACTGTTGGCACTCCATCTG) and Rec-kh-rv2 (GGGCTTCAACAACTGAAAAG) were used to amplify a 2190 bp region spanning the cut site of gRNA-sw4, while primers eGFP-fw (AAGTCGTGCTGCTTCATGTG) and Vasa-rv (GTAAAAGCCGCATTTTCCAA) were used to amplify a 2303 bp region across the cut site of gRNA-sw3. Gene amplification reactions were performed using Phusion® High-Fidelity PCR Master Mix (New England Biolabs). Sanger sequencing (Genewiz, San Diego) with primers Rec-kh-rv2 and eGFP-fw was used to confirm the sequence of the integration sites.

Primary drive transmission assessments

Drive transmission and HDR conversion rates through the male and female lineages were assessed in sequential en masse outcrosses of Reckh individuals to WT. Each cross comprised 30 females and 15 males and was performed in three replicate cages. A representative subset of the progeny of each cross was scored for the presence of the GFP fluorescent marker and the eye color phenotype (black, white, or mosaic) in adults. A schematic of the crossing performed is reported in Fig. 2.

Drive transmission is defined as the percentage of individuals inheriting the Reckh element. Gene conversion or HDR rate is defined as the percentage of kh alleles converted to Reckh by HDR copying and is calculated using the formula [2(X − 0.5n)/n] (“X” is the number of GFP+ individuals and “n” the total number of mosquito counted)8.

Female fecundity and fertility

Homozygous Reckh (khRec+/khRec+), heterozygous Reckh carrying a copy of the drive and a kh allele (khRec+/kh), WT (kh+/kh+), and white-eye (kh/kh)9 females were included in this analysis. Adult females 5–7 day old were offered a blood meal for 45 min over 2 consecutive days and unfed females removed. After 3 days, single females were set up to oviposit in 16 oz (~454 cm3) paper cups containing a plastic oviposition cup lined with damp filter paper. Eggs were counted the next day using a stereomicroscope and transferred to water cups lined with filter paper for hatching. Larvae emerging from single egg batches were counted at the first or second instar (L1–L2). Fecundity refers to the number of eggs laid by a single female and fertility to the proportion of larvae hatching from these individual egg batches. A one-way ANOVA with Tukey’s multiple comparison post hoc test was used to assess significant differences (p > 0.05) in the performance of females from the four groups tested.

Male contribution to the following generation

Triplicate cages were seeded with 75 Reckh homozygous males, 75 WT males, and 150 WT females. All individuals were added to the cage as 3–7-day-old adults and females were offered a blood meal over two consecutive days. Approximately, 2000–2500 L4 larvae were selected randomly from the progeny of each replicate cage and scored for the presence of GFP. A two-tail binomial test was used to compare the observed and expected distributions and test for significant (p > 0.05) deviation from equal frequency (50%) of GFP+ and GFP individuals.

Cage trial set up and maintenance

A schematic representation of the cage trial protocol implemented is shown in Supplementary Fig. 2. The trial consisted of 18 nonoverlapping generations and was conducted in 5000 cm3 cages essentially as described by Pham et al.9. Triplicate cages (A–C) were seeded with three single release ratios, 1:1, 1:3, 1:9, of 3–5 day old age-matched Reckh heterozygous to WT (Reckh:WT) male adults (100 in total), and 100 WT adult females were added to reach an equal sex ratio for a total of 200 individuals per cage. The number of Reckh males was 50 in the 1:1 release cages, 25 in the 1:3 cages, and 10 in the 1:9 cages. Adults 5–7 days old were offered a blood meal over two consecutive days. Three days later, dead adults were removed from each cage and an egg cup was provided for two days. Of the hatching larvae: 200 L1–L2 were selected at random and reared to adulthood to establish the following generation, 500 L1–L2 were selected randomly to assess the progression of the drive by screening the eye phenotype in L4 larvae and pupae (see “Screening of the eye phenotype” section), and the remainder reared to L4 and stored in ethanol for population counts and molecular analysis. Due to the small initial number of transgenics, the only exception to the random selection of the 200 L1s to seed the following generation was that individuals from generation 1 of the 1:9 cages were all screened for their eye phenotype and new cages were seeded with the same proportion of drive individuals found. Finally, the screening of cage 1:3B was carried out for additional two generations for a total of 20 generations.

Screening of the eye phenotype

A sample of 500 randomly selected L4 larvae were scored at each generation for the presence of GFP fluorescence (GFP+ and GFP) and separated in two corresponding larval trays; pupae emerging from each tray were screened for eye color (black, white, or mosaic). The phenotypes were reported as follows: GFP+/kh+ (drive individuals with black eyes); GFP+/kh (drive individuals with white eyes); GFP+/khmos (drive individuals with mosaic eyes); GFPkh+ (non-drive individuals with black eyes); GFP/kh (non-drive individuals with white eyes); GFP/khmos (non-drive individuals with mosaic eyes). Among these, individuals with phenotypes GFP+/kh and GFP/kh were stored as adults at −20 °C for sequencing.

Population count

L4 larvae collected throughout the experiments were stored in 50 mL conical centrifuge tubes filled with ethanol. Before counting, ethanol was rinsed off and larvae were re-suspended in a fixed volume of deionized water and placed onto a shaker moving at a constant speed. Larvae were collected using a fixed-volume scoop and counted before returning them to the shaker. A total of 6–9 measurements were taken per cage every two generations. The estimated population size was calculated by averaging the number of larvae from replicate measurements and multiplying by the conversion factor (volume of water/scoop volume). The only exception to this method of counting was generation 1 of the 1:9 cages where the whole L4 population was counted.

Sanger sequencing on single mosquitoes

Genomic DNA was extracted from whole single adult mosquitoes using either the Wizard Genomic DNA Purification Kit (Promega) or the DNeasy Blood & Tissue Kit (Qiagen). All gene amplification reactions were performed using the Phusion High Fidelity PCR Master Mix with HF Buffer (New England Biolabs). To analyze the non-drive allele, primers KhE5-4 (GACGGTGACACTGTTCATGC) and KhE5-3 (CAGATGGCATGTGCATCCTC) were used to generate a 372 bp amplicon spanning the gRNA-directed cut site in the kh gene. Sanger sequencing (Genewiz, San Diego) of the non-drive amplicons was performed using primer KhE5-4. To analyze the Reckh drive allele, primers KhE4 (CGTTCGAGTAGCACGTTG) and Agam3 rv (CAGGTGTAGAAGAAAACACGTTG) were used to produce a 1287 bp amplicon. Sanger sequencing (Genewiz, San Diego) of the Reckh amplicon was performed using primer KhE4. Sequencing results from mixed traces were resolved using CRISP-ID (

DNA extraction and amplification from pooled mosquitoes

Genomic DNA from individuals used to seed the 1:3B cage (generation 0) was extracted from pools of 20 adults (total of ~140); while DNA from individuals from the same cage at generations 8 and 14 was extracted from pools of 50 larvae (total of 300 each). Extractions were performed using the DNeasy Blood & Tissue Kit (Qiagen) according to manufacturer’s protocol with an overnight initial lysis step. An equal volume of genomic DNA was pooled from each replicate extraction and used as template for amplification. Gene amplification was performed using the Phusion High Fidelity PCR Master Mix with HF Buffer (New England Biolabs) and primers KhE5-4 (GACGGTGACACTGTTCATGC) and KhE5-3 (CAGATGGCATGTGCATCCTC). Generated amplicons were purified from 1% agarose gels using the Zymoclean Gel DNA Recovery Kit (Zymo Research) before library preparation.

Library preparation and sequencing

Illumina libraries were prepared for each of three samples (G0, G8, and G16 from cage 1:3B) using the NEXTFLEX PCR‐free library preparation kit and NEXTFLEX Unique Dual Index Barcodes (BIOO Scientific) following the manufacturer’s instructions. The input amount of DNA was 500 ng. The ends of the DNA were repaired and adenylated. The reaction was cleaned using AMPure XP magnetic beads and Illumina barcoded adapters were ligated onto the blunt-end adenylated product. The adapter-ligated product was cleaned using AMPure XP beads. DNA quantity was measured by Qubit DNA HS assay and the fragment size assessed by Agilent Bioanalyzer 2100 DNA HS chip assay at the genomics facility of the University of Utah (GNomEx) where the libraries were sequenced on the Illumina NovaSeq with the SP flowcell 2 × 250 paired end.

Sequencing data analysis

The raw paired-end Illumina reads from the amplified genomic region were cleaned for low quality and trimmed for the presence of adapters using Trimmomatic v0.3546. High-quality reads were mapped against the amplicon sequence using BWA-MEM v0.7.847 and the alignments sorted using SAMtools v1.948. Mapped paired-end reads were extracted using Picard Tools v1.96 (, and then joined to reconstruct the complete amplicon sequence using PEAR v0.9.849. Identical amplicon sequences were clustered using module fastx_collapser in FASTX-ToolKit v0.0.14 ( Clustered sequences were aligned to the reference amplicon with MAFFT v750 under FFT-NS-2 tree-based progressive method with 1PAM/K = 2 scoring matrix, if they were represented by ≥3 paired-end reads in each dataset. After alignment with the amplicon, the analysis was focused on a 34 bp target sequence including the 23 bp of the gRNA to identify single-nucleotide polymorphisms and/or INDELs in this region. The final quantification of mutations at the target site was measured as relative frequency of paired reads in sequence variants represented in at least 100 reads.

Functional resistant allele assessments

Individuals carrying a mutated functional kh allele (kh+R) due to the presence of a CAG > GCA-Q330A substitution affecting the PAM site were isolated from non-drive black-eyed (GFP/black) individuals from cage 1:3B at generation G16. Resistance of the kh+R allele to Cas9-induced cleavage was assessed in the progeny of the cross between males heterozygous for a copy of the Reckh drive allele and the kh+R allele (khRec+/kh+R) to WT females by scoring the frequencies of GFP+ and GFP mosquitoes (expected to be ~50% in case of resistance to cleavage).

Allele competition experiments were conducted in four replicate cages (A–D) each seeded with 200 individuals heterozygous for a copy of the Reckh drive allele and a copy of the kh functional resistant allele (khRec+/kh+R) with a 1:1 sex ratio. Allele competition was inferred from the eye phenotype of the progeny of these crosses by scoring for the presence of the GFP fluorescent marker (GFP+ or GFP) and the eye color (black, white, or mosaics) in adults. This was carried out for six discrete (nonoverlapping) generations by screening a representative sample of ~300 adults at each generation and seeding new cages with 200 randomly picked individuals, as described for the gene-drive cage trials. In this set-up, assuming random mating, equally competitive modified kh alleles are expected to maintain a phenotypic ratio of 3:1 GFP+:GFP individuals in each generation; deviations from this ratio would signify unequal competitiveness.

Modeling of cage population dynamics

Empirical data from the nonoverlapping gene-drive experiments were used to parameterize a model of Cas9/gRNA-based homing gene-drive including resistant allele formation, and a stochastic implementation of the fitted model was used to qualitatively compare the time series of observed genotype frequencies to model-predicted ones. Model fitting was carried out for all nine gene-drive cage experiments using Markov chain Monte Carlo methods in which estimated parameters related to loads, resistant allele generation, and the consequences of maternal deposition of Cas9/gRNA complexes were used.

We considered discrete generations, random mixing, and Mendelian inheritance rules at the gene-drive locus, with the exception that for adults heterozygous for the homing allele (denoted by “H”) and WT allele (denoted by “W”), a proportion, c, of the W alleles are cleaved, while a proportion, 1 − c, remain as W alleles. Of those that are cleaved, a proportion, pHDR, are subject to accurate HDR and become H alleles, while a proportion, (1 − pHDR), become resistant alleles. Of those that become resistant alleles, a proportion, pRES, become in-frame, functional, cost-free resistant alleles (denoted by “R”), while the remainder, (1 − pRES), become out-of-frame, nonfunctional, or otherwise costly resistant alleles (denoted by “B”). The value of pHDR is allowed to vary depending on whether the HW individual is female or male, and values for female- and male-specific HDR parameters were estimated based on G0 crosses that provided direct information on them.

The effects of maternal deposition of Cas9/gRNA complexes were accommodated after computing the gene-drive-modified Mendelian inheritance rules. If offspring having a W allele had a mother having the H allele, then this would lead to Cas9/gRNA complexes being deposited in the embryo by the mother, possibly resulting in cleavage of the W allele. We considered cleavage to occur in a proportion, pMC, of these embryos, with a proportion, pMR, of the cleaved W alleles becoming R alleles, and the remainder, (1 − pMR), becoming B alleles.

These considerations allow us to calculate expected genotype frequencies in the next generation, and to explore the impacts of loads and maternal deposition parameters that maximize the likelihood of the experimental data. Estimated parameters include loads in females associated with having one or two copies of the H allele or the BB genotype, and pRES, pMC, and pMR, as defined earlier. A stochastic version of the fitted model was implemented using a discrete generation version of the Mosquito Gene-drive Explorer (MGDrivE) model51 with an adult population size of 200. The complete modeling framework is described in the “S1 Text” section of Pham et al.9.

Statistical analysis

Statistical tests were performed as detailed in the relevant method sections using GraphPad Prism 8.4.2.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.