ArticlePDF Available

Abstract and Figures

The recent development of a CRISPR-Cas9-based homing system for the suppression of Anopheles gambiae is encouraging; however, with current designs, the slow emergence of homing-resistant alleles is expected to result in suppressed populations rapidly rebounding, as homing-resistant alleles have a significant fitness advantage over functional, population-suppressing homing alleles. To explore this concern, we develop a mathematical model to estimate tolerable rates of homing-resistant allele generation to suppress a wild population of a given size. Our results suggest that, to achieve meaningful population suppression, tolerable rates of resistance allele generation are orders of magnitude smaller than those observed for current designs for CRISPR-Cas9-based homing systems. To remedy this, we theoretically explore a homing system architecture in which guide RNAs (gRNAs) are multiplexed, increasing the effective homing rate and decreasing the effective resistant allele generation rate. Modeling results suggest that the size of the population that can be suppressed increases exponentially with the number of multiplexed gRNAs and that, with four multiplexed gRNAs, a mosquito species could potentially be suppressed on a continental scale. We also demonstrate successful proof-of-principle use of multiplexed ribozyme flanked gRNAs to induce mutations in vivo in Drosophila melanogaster – a strategy that could readily be adapted to engineer stable, homing-based drives in relevant organisms.
Content may be subject to copyright.
1
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
www.nature.com/scientificreports
Overcoming evolved resistance to
population-suppressing homing-
based gene drives
John M. Marshall1, Anna Buchman2, Héctor M. Sánchez C.
3 & Omar S. Akbari2
The recent development of a CRISPR-Cas9-based homing system for the suppression of Anopheles
gambiae is encouraging; however, with current designs, the slow emergence of homing-resistant alleles
is expected to result in suppressed populations rapidly rebounding, as homing-resistant alleles have
a signicant tness advantage over functional, population-suppressing homing alleles. To explore
this concern, we develop a mathematical model to estimate tolerable rates of homing-resistant allele
generation to suppress a wild population of a given size. Our results suggest that, to achieve meaningful
population suppression, tolerable rates of resistance allele generation are orders of magnitude smaller
than those observed for current designs for CRISPR-Cas9-based homing systems. To remedy this, we
theoretically explore a homing system architecture in which guide RNAs (gRNAs) are multiplexed,
increasing the eective homing rate and decreasing the eective resistant allele generation rate.
Modeling results suggest that the size of the population that can be suppressed increases exponentially
with the number of multiplexed gRNAs and that, with four multiplexed gRNAs, a mosquito species
could potentially be suppressed on a continental scale. We also demonstrate successful proof-of-
principle use of multiplexed ribozyme anked gRNAs to induce mutations in vivo in Drosophila
melanogaster – a strategy that could readily be adapted to engineer stable, homing-based drives in
relevant organisms.
e concept of using homing-based gene drive systems to rapidly invade wild populations and spread eector
genes (e.g. conferring pathogen resistance) or to suppress and eliminate populations was rst suggested by Burt in
20031. ese systems have the remarkable ability to cheat during meiosis, enabling them to rapidly spread into a
population even if they confer a tness cost to their host (2, 3). ey achieve this by encoding a sequence-specic
nuclease that generates a double-stranded break at one or more specic target loci in a host’s genome, directly
opposite the drive. To survive, the cell is forced to rapidly repair the DNA break using its endogenous DNA repair
machinery. Repair of the break using the homology-directed repair (HDR) pathway, for instance, can result in
the drive system being perfectly copied into its competing allele. When this occurs in a germline cell, it eectively
converts a heterozygote into a homozygote, allowing the system to circumvent traditional Mendelian inheritance
patterns and to drive into a population2, 3. e rst decade following this proposition saw moderate progress in
the development of homing-based drive systems in the African malaria vector, Anopheles gambiae2, 4, 5. More
recently, the development of the CRISPR-Cas9 system has unlocked enormous potential for this technology, with
highly ecient drive systems being developed in quick succession to modify laboratory populations of Drosophila
melanogaster6, Saccharomyces cerevisiae7, the Asian malaria vector, Anopheles stephensi8, and the main African
malaria vector, An. gambiae9.
e homing-based drive systems developed using CRISPR-Cas9 have a number of highly desirable features:
they are relatively straightforward to adapt to new target sequences and to port to other species, and the constructs
engineered thus far have extremely high transmission rates, being inherited by 90–99% of the ospring of hete-
rozygous parents69. However, these systems are not without their shortcomings. Firstly, the CRISPR-Cas9-based
constructs engineered to date are associated with high tness costs, although this is likely not an inherent problem
with CRISPR-Cas9 systems, and secondly, the homing process has been shown to be highly error-prone, leading
1Divisions of Biostatistics and Epidemiology, School of Public Health, University of California, Berkeley, CA, 94720,
USA. 2Department of Entomology, Center for Disease Vector Research, and Institute for Integrative Genome
Biology, University of California, Riverside, CA, 92521, USA. 3Bioinformatics Research Group, School of Medicine,
Tecnológico de Monterrey, Estado de México, 52926, México, USA. Correspondence and requests for materials
should be addressed to J.M.M. (email: john.marshall@berkeley.edu) or O.S.A. (email: omar.akbari@ucr.edu)
Received: 12 December 2016
Accepted: 18 April 2017
Published: xx xx xxxx
OPEN
www.nature.com/scientificreports/
2
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
to the creation of mutant alleles within a few generations8, 9. A signicant number of these mutant alleles consist
of small deletions that preserve the reading frame and could represent homing-resistant alleles, i.e. mutant alleles
that are not associated with a tness cost8, 9. is latter shortcoming is of concern for population suppression
strategies because homing-resistant alleles have a strong selective advantage over functional homing alleles that
disrupt an essential gene, leading to suppressed populations rapidly rebounding.
Mutant alleles may be generated in several ways. For instance, they can emerge when the cell mends DNA
damage at the target site using the non-homologous end-joining (NHEJ) pathway instead of HDR following
drive-induced target site cleavage. Mutant alleles may also arise due to incomplete or imperfect copying during
HDR. e CRISPR-Cas9 system is particularly vulnerable to this due to its large size – the system consists of pro-
moters, the Cas9 gene, guide RNAs and, depending on the strategy being implemented, multiple eector genes
and associated regulatory elements, all of which need to be perfectly copied during HDR to ensure spread into a
population. Indeed, for the CRISPR-Cas9 homing construct engineered in An. gambiae9, incomplete homing or
internal deletion events were observed in 47% (15 out of 32) of screened organisms in which an errorless homing
event was not observed. ese were derived from a minimum of seven originating erroneous homing events, one
of which produced a small deletion that preserved the reading frame and may be considered a homing-resistant
allele. Homing-resistant alleles may also arise de novo via random target site mutagenesis, and some organisms
may intrinsically be resistant to homing activity at a given site due to genetic variation within a species.
erefore, while CRISPR-Cas9-based homing systems have enormous potential for the targeted engineering
of populations, signicant technical improvements are required if this technology is to be successfully imple-
mented in the eld2, 10. Here, we focus specically on the issue of homing-resistant allele generation for popu-
lation suppression homing systems. We largely ignore homing allele tness costs in this analysis as we consider
these to be surmountable through tailored engineering eorts – in one of the constructs engineered thus far, t-
ness costs seem to result from the transgene being inserted into an eye color gene8, and in another, due to the ele-
ment copying itself to somatic as well as germline cells9, both of which we believe to be addressable. e impact of
homing-resistant alleles on homing-based population replacement strategies has been described by Noble et al.11
along with a design strategy that selects against the resistant alleles; however, this solution does not apply to the
population suppression systems that we explore here.
To address the impact of resistant alleles on homing-based population suppression systems, we develop a
mathematical model to estimate the maximum tolerable resistant allele generation rates to achieve stable,
long-term suppression for populations of various sizes. Our results suggest that, to achieve meaningful popula-
tion suppression, tolerable rates of resistant allele generation are orders of magnitude lower than those observed
for current CRISPR-Cas9-based homing systems. We describe how the required rates can be achieved by target-
ing multiple locations in a gene through guide RNA (gRNA) multiplexing2, 3, 11, 12. Furthermore, we demonstrate
successful adaptation of a multiplexed ribozyme-gRNA-ribozyme (RGR) approach previously demonstrated
in yeast13 and in mammalian tissue culture cells13, 14 to in vivo mutagenesis of target genes in Drosophila mel-
anogaster, and discuss possible designs for, and challenges inherent in, the gRNA multiplexing approach for
engineering stable, homing-based suppression gene drive systems. Finally, we explore the scale of population
suppression that can be achieved by using this approach.
Results
e homing population suppression system we explore here is based on that described by Hammond et al.9 in
which the CRISPR-Cas9 system is designed to target a gene required for female fertility. is has the eect that
females homozygous for the homing allele are infertile; however, heterozygous and wild-type females have at
least one functional copy of the fertility gene and are fertile. For suciently high homing rates and small tness
costs, this system can spread into a population while it reduces population fertility, eventually leading to a popu-
lation crash1. Hammond et al.9 describe three strains that they engineered with this design. We consider the most
successful of these – construct 7280 – for which the transmission rate from heterozygotes was ~99%, and hete-
rozygous females had their tness reduced by 90.7%. Approximately half (~47%) of those who did not inherit a
functional homing allele from a heterozygous parent inherited a copy with errors, and a seventh of the erroneous
homing events that produced these alleles generated alleles that could be considered homing-resistant. Although
the Hammond et al.9 construct was not proposed as a useable drive system, we consider its parameter values as a
basis that future homing-based drive systems could build upon.
Model framework. e framework used to model this system is described in the Materials and Methods; but
in short, we denote the homing allele as “H”, the wild-type allele as “h”, and the homing-resistant allele (mutant
allele with no associated tness cost) as “R. HH females are infertile, while all other genotypes are fertile. Hh
males and females produce H gametes in the germline at a frequency equal to (1 + e)/2, where e denotes the e-
ciency of homing, or “homing rate. Hh individuals also produce R gametes in the germline at a frequency equal
to ρ/2, where ρ denotes the resistant allele generation rate. e remaining gametes produced by Hh individuals
are wild-type and are produced at a frequency equal to (1 e ρ)/2. Females heterozygous for the homing allele
have their fertility reduced by a fraction, s, while other genotypes are equally fertile (except for HH females, which
are infertile). e crosses describing this system are shown in Supplementary Figure1.
We use a discrete population, stochastic framework incorporating density-dependence at the larval stage to
model this system. Our framework is modied from one previously used to examine the spread of homing endo-
nuclease genes (HEGs) through populations of An. gambiae5, the main malaria vector and the species in which
the CRISPR-Cas9-based constructs were developed by Hammond et al.9. is framework incorporates the egg,
larval, pupal and adult life stages. Generations are overlapping and adult females mate once, retaining the genetic
material of the male they mate with for the duration of their adult life. Since we are modeling a population sup-
pression system, a discrete, stochastic model is needed to capture the chance events that happen at low population
www.nature.com/scientificreports/
3
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
sizes. is also enables the simulation of a population crash. Density-dependence at the larval stage is important
to include as it captures the phenomenon in which more larvae survive to emergence when populations are small
due to reduced larval competition. To model this, we consider a monotonic increase in larval mortality with larval
density, as applied by Deredec et al.5.
Expected dynamics of present constructs. With the modeling framework established, we explore the
predicted dynamics of construct 7280, the best-performing construct engineered by Hammond et al.9, in a popu-
lation size of N = 10,000 adult mosquitoes. e results described in Fig.1 correspond to a homing rate of e 98%
(2 × (99–50%)) and a resistant allele generation rate of ρ 0.13% ((1 e) × 15/32 × 1/7), i.e. of the wild-type
alleles from heterozygotes that were not converted to homing alleles, 15/32 were mutant alleles and 1/7 of the
events that produced these resulted in what may be considered homing-resistant alleles. e scenario in which
females heterozygous for the homing allele have their fertility reduced by 90.7% is shown in Fig.1A. Here, we see
that gene drive occurs slowly and population suppression is moderate and transient. e total adult population
falls by ~64% approximately one and three-quarter years following a 1:1 seeding release of HH males to hh males
and females; however, this suppression is short-lived – a population reduction of 50% or more is only maintained
for about ve and a half months before the population rebounds.
Henceforth, let’s imagine that the fertility costs of the homing allele in heterozygous females can be prevented
through engineering eorts to ensure that the CRISPR-Cas9 system is only expressed in germline cells. is sce-
nario is shown in Fig.1B. Here, we see that gene drive and population suppression occur more quickly, and the
extent of population suppression is slightly greater (the population is suppressed by ~69% at its peak). However,
the duration of suppression is very short – a population reduction of 50% or more is only maintained for about
two months. is is far less than what is hoped for gene drive-based population suppression strategies, and is a
consequence of the quick emergence of homing-resistant alleles once the gene drive system becomes prevalent in
the population, leading to a population rebound.
Design criteria for population elimination. e constructs engineered by Hammond et al.9 are clearly
inadequate to lead to meaningful population suppression for population sizes of 10,000 adult mosquitoes; but
presumably if the homing rate were increased and/or the resistant allele generation rate were decreased, then it
may be possible to eliminate a specic population. In Fig.2, simulations are shown in which homing eciency
is maintained at 98% while the resistant allele generation rate is reduced from 0.1% (103) to 0.001% (105) to
0.00001% (107) for populations of 1,000, 10,000 and 100,000 adult mosquitoes. Here, we see that, for a resistant
allele generation rate of 103 (Fig.2A–C), which is approximately what was observed for the Hammond et al.9
construct, we do not expect to be able to meaningfully suppress an adult population even as small as 1,000 indi-
viduals. However, if we reduce the resistant allele generation rate by two orders of magnitude to 105 (Fig.2D–F),
then we expect to eliminate populations of sizes 1,000 and 10,000, but not a size of 100,000. As the resistant allele
generation rate is further reduced by an additional two orders of magnitude to 107 (Fig.2G–I), we expect to
eliminate adult populations of all sizes up to 100,000.
is trend of being able to eliminate populations of larger size with smaller resistant allele generation rates is
intuitive as, in a larger population, there are more opportunities for error-prone homing events to occur, leading
to the emergence of homing-resistant alleles. ese resistant alleles will quickly be selected for, reversing any prior
population suppression. e design target for the resistant allele generation rate will therefore be determined by
the population size we wish to eliminate.
Figure 1. Predicted population dynamics for the present CRISPR-Cas9-based population suppression homing
constructs. Here we model the predicted dynamics of the best-performing construct engineered by Hammond
et al.9 in a population of 10,000 adult mosquitoes. e homing rate for this construct is ~98% and the resistant
allele generation rate is ~0.13%. e model framework is described in the Materials and Methods. In panel (A),
the dynamics are shown for the scenario in which females heterozygous for the homing allele have their fertility
reduced by 90.7%. In panel (B), the same construct is modeled in the absence of a fertility cost. In both cases,
population suppression is moderate and short-lived due to the generation of homing-resistant alleles leading
to a population rebound. Red lines represent individuals having at least one copy of the homing allele (i.e.
genotypes Hh, HR and HH), green lines represent individuals having at least one copy of the homing-resistant
allele (i.e. genotypes hR, HR and RR), and blue lines represent the total population. Solid lines represent the
median population size for 100 repetitions of the stochastic model, while shaded regions represent the 25–75%
quartile range in these simulations, where this diers signicantly from the median.
www.nature.com/scientificreports/
4
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
e homing rate, however, does not factor into these design considerations, at least for already high homing
rates. Figure3 shows the probability of population elimination as homing eciency is varied between 97% and
99.99% and the resistant allele generation rate is varied between 102 and 107 for a population of 10,000 adult
mosquitoes. e elimination probability is calculated as the proportion of simulations (from a total of 100 per
parameter set) in which the An. gambiae population was eliminated within 950 days of a 1:1 release of HH males
to hh males and females. Here we see that, for a population size of 10,000 adults, population elimination is highly
likely for resistant allele generation rates smaller than 105, and is unlikely for rates above 104. ere is a critical
rate between these two values at which the population is equally likely to either rebound or be eliminated and,
interestingly, these dynamics are independent of the homing rate for e > 98%.
Dependence of the design criteria for ρ on N. e independence of elimination probability and homing
rate (for already high homing rates) means that we can focus our attention on achieving a resistant allele gener-
ation rate, ρ, small enough such that population elimination is likely for a given population size, N. Stochastic
simulations become highly computationally intensive as population size increases, and so we seek a relationship
between N and the corresponding resistant allele generation rate, ρ, for which we can be 90% sure of achieving
population elimination (or sure with some other degree of certainty). To this end, Fig.4A depicts elimination
probability as a function of ρ as we vary N between 1,000 and 100,000. e familiar case of a population size of
10,000 is shown in light gold and indicates that elimination is ~10% likely for a ρ value of 104 and ~80% likely
for a ρ value of 105. As the population size increases from 1,000 to 100,000, we see that the ρ value required to
achieve an elimination probability of 90% or higher becomes signicantly smaller.
e form of the relationship between N and ρx, the resistant allele generation rate leading to an elimination
probability of x, is depicted in Fig.4B for selected elimination probabilities. Fortunately for our ability to extrap-
olate to larger population sizes, there is a linear relationship between 1/N and ρx for all elimination probabilities
investigated, i.e.:
ρ=.cN/(1)
x
x
Here, cx represents the slope from Fig.4B corresponding to the elimination probability, x. To be 90% sure of pop-
ulation elimination, cx is 0.0410, to be 95% sure, cx is 0.0199, and to be 99% sure, cx is 0.00391. is means that, to
be 90% sure of eliminating a population of size 1,000, ρ should be less than 4.1 × 105 (0.0041%), and to be 90%
sure of eliminating a population of size 100,000, ρ should be less than 4.1 × 107 (0.000041%). To be 95% sure of
Figure 2. Homing and resistant allele trajectories for a range of population sizes and resistant allele generation
rates. Here, we model a population suppression homing construct with a homing rate of 98% and no fertility
cost. In panels (A–C) the resistant allele generation rate is 0.1% (103), in panels (D–F) it is 0.001% (105), and
in panels (G–I) it is 0.00001% (107). In the lemost panels (A,D and G), a population of 1,000 is modeled,
in the middle panels (B,E and H), it is 10,000, and in the rightmost panels (C,F and I), it is 100,000. Red lines
represent individuals having at least one copy of the homing allele, green lines represent individuals having at
least one copy of the homing-resistant allele, and blue lines represent the total population. Solid lines represent
the median value obtained from 100 repetitions of the stochastic model, while shaded regions represent the
25–75% quartile range, where this diers signicantly from the median. As the resistant allele generation rate is
reduced, we expect to eliminate populations of larger sizes.
www.nature.com/scientificreports/
5
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
eliminating populations of these sizes, ρ should be about half that predicted for a 90% chance of elimination, and
for a 99% chance of elimination, ρ should be an order of magnitude smaller than that predicted for a 90% chance
of elimination. ese estimates are derived for a mosquito population growth rate of RM = 9.1 per generation,
based on mosquito count data from the Garki project in Nigeria15. We repeated these analyses for low, medium
Figure 3. Dependence of population elimination probability on homing rate and resistant allele generation
rate. Here, we model a population suppression homing construct in a population of 10,000 adult mosquitoes.
Each pixel represents a combination of homing and resistant allele generation rates for which the simulation
was run. Pixel shadings represent the proportion of 100 simulations in which population elimination was
achieved within 950 days of a seeding 1:1 release of HH males to hh males and females. Both rate parameters
were sampled logarithmically to gain higher resolution at high homing rates and low resistant allele generation
rates. e white region represents impossible combinations of rate parameters (the rates would sum to >1).
Population elimination probability is independent of the homing rate (for already high homing rates) and
critically dependent on the resistant allele generation rate.
Figure 4. Relationship between population size and the resistant allele generation rate required for a given
population elimination probability. (A) Elimination probability as a function of resistant allele generation rate
for a range of population sizes, N, between 1,000 and 100,000. Sigmoidal curves are tted to data points covering
30 resistant allele generation rates sampled logarithmically between 102 and 107. (B) Linear relationship
between 1/N and the resistant allele generation rate leading to a given probability of population elimination.
Values of 1/N are as shown in panel A, and resistant allele generation rates are inferred from the sigmoid curves.
Faint lines in both panels represent interpolation between simulated data points while solid lines represent
tted linear relationships. ere is a clear linear relationship between 1/N and the resistant allele generation rate
leading to a given elimination probability.
www.nature.com/scientificreports/
6
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
and high population growth rates (RM = 2, 6 and 12 per generation, respectively) and found that, while the slope,
cx, changes, the linear relationship in Equation1 still holds (Supplementary Figure2).
e linear nature of the relationship between 1/N and ρx is understandable since each mosquito presents
an opportunity for a resistant allele to emerge and prevent population elimination; however, it is reasonable to
question whether population structure has a signicant eect on the tolerable resistant allele generation rate.
Supplementary Figure3 explores the relationship between 1/N and ρx for randomly mixing populations of sizes
10,000–50,000 and for 1–5 randomly mixing populations each of size 10,000 that exchange migrants with the
other populations at a rate of 1% per adult mosquito per generation. e linear relationship between 1/N and ρx is
unchanged by the presence of population structure, which supports the intuition that each mosquito presents an
opportunity for resistant allele emergence independent of population structure.
e population size that we wish to eliminate will vary depending on our goals; but one proposition for
homing-based gene drive has been to eliminate a disease vector species such as An. gambiae on a continental
scale. Assuming there are about ten times as many adult An. gambiae mosquitoes on the African continent as
there are people, this suggests a population of ~10 billion (1010). e eective population size for a given elimi-
nation probability will be smaller than this due to population structure details and seasonal uctuations in An.
gambaie population size16; however, we consider this larger population size to provide conservative estimates of
required ρ values. To be 90% sure that resistant alleles will not prevent elimination of a population this size, ρ
should be less than 4.1 × 1012 (0.0000000004%). To be 95% sure, ρ should be less than 2.0 × 1012, and to be 99%
sure, ρ should be less than 3.9 × 1013.
Multiplexing gRNAs. e ρ values required to prevent resistant alleles from interfering with the elimination
of an An. gambiae population on the scale of the African continent are vanishingly small; but interestingly, the ρ
value required to have a 90% chance of suppressing a population of just 1,000 adult mosquitoes is already several
orders of magnitude smaller than that observed for the best-performing construct of Hammond et al. 9. Given the
inevitability of the evolution of homing-resistant alleles, mitigating their impact is imperative to creating func-
tional and stable homing-based population suppression gene drive systems.
A promising strategy for achieving this feat is the multiplexing of gRNAs in the gene drive to target multiple
sequences. is idea has been previously proposed to increase the stability of drives2, 3, 11, 17; however, to date
only one multiplexing strategy has been demonstrated to function in a whole-animal model12. erefore, to fur-
ther expand the toolbox for multiplexing gRNAs in whole animals, here we test the eectiveness of a technique
previously demonstrated in yeast13 and in mammalian tissue culture cells14 that relies on anking gRNAs with
self-cleaving ribozymes, known as the ribozyme-gRNA-ribozyme (RGR) approach in vivo in Drosophila melano-
gaster13. To validate the ability of this technique to induce mutations in vivo in Drosophila, we generated plasmid
OA-16 that contains two multiplexed RGRs, the rst targeting the white gene and the second targeting the yellow
gene at target sequences previously validated18, driven by a single polymerase-2 ubiquitin promoter19 (Fig.5A,B).
e plasmid also contains a white gene that could be targeted by the white gRNA and used for detecting trans-
genic individuals bearing the OA-16 construct.
is plasmid was integrated site-specically into the Drosophila attP line BSC (Bloomington Stock Center)
24486. Generated transgenic males and females were individually mated to females and males of transgenic line
BSC 51324 (vasa-Cas9), and the progeny of the resulting crosses were scored (Fig.5C). As expected, transformant
ies bearing the OA-16 plasmid had no mutations in the white or yellow genes in the absence of Cas9 (Fig.5D,
le). All the scored ospring (712 from OA-16 male/Cas9 female crosses and 1053 from OA-16 female/Cas9
male crosses) had completely white or signicantly variegated eyes (Fig.5D), indicating that the rst of the two
RGRs had a relative somatic mutation eciency near 100%, at least as measured by the proportion of individu-
als in which mutation events occurred. Although precise data were not compiled, mutation-bearing individuals
gave rise to completely mutated, i.e., white-eyed, progeny at very high frequency, indicating functional germline
transmission which is important for development of gene drives. Additionally, 87% +/ 0.41 (616/712) of the o-
spring of OA-16 male/Cas9 female crosses and 60.5% +/ 0.43 (637/1053) of the ospring from OA-16 female/
Cas9 male crosses had a predominantly yellow cuticle (Fig.5D, right), indicating that the second RGR was also
functional, albeit with a signicantly lower eciency than the rst. It should be noted that the eciency of the
second gRNA may be underestimated, as small mutant patches of yellow cuticle tissue are generally more dicult
to detect than white patches ofommitidia facets in the compound eye. e presence of mutations was conrmed
by sequencing of PCR products that span the cleavage site (Fig.5E). Together, these data conclusively provide a
proof-of-principle for the feasibility of the RGR approach as a method for multiplexing gRNAs in whole animals.
Design requirements for multiplex number. This demonstration of the multiplexing of gRNAs is
encouraging for the same being achieved in insect species that transmit human diseases, such as An. gambiae.
Multiplexing is expected to increase the eective homing rate, as only one of several target sites must have a
functional copy of the homing allele to be capable of homing. at said, as depicted in Fig.3, the probability of
population elimination is independent of the homing rate for e > 98%, but is highly dependent on the resistant
allele generation rate, ρ. We derive the eective resistant allele generation rate for two and three multiplexed
gRNAs in Supplementary TextS1 and nd that this is approximately equal to ρ2 and ρ3, respectively. is logically
follows since resistant allele generation in the presence of multiplexing requires all gRNA target sites to have a
homing-resistant allele. In Supplementary TextS1 we show that, although homing-resistant alleles may accumu-
late in a composite allele with multiple target sites, partially resistant composite alleles are rarely generated and
are frequently converted to homing alleles in subsequent generations. e rate of completely resistant composite
alleles emerging is therefore approximately equal to the rate of resistant alleles emerging at all target sites at once,
i.e. ρm for a multiplex number of m.
www.nature.com/scientificreports/
7
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
e eective resistant allele generation rate therefore becomes exponentially smaller as the number of multi-
plexed gRNAs increases. For a baseline ρ value of 0.13%, the eective ρ value becomes ~1.7 × 106 for a multiplex
number of two, and ~2.2 × 109 for a multiplex number of three. Following on from Equation1, for a baseline
resistant allele generation rate of ρ at a single site, the multiplex number, m, we must achieve to have a chance x of
eliminating a population of size N is given by:
ρ>− .mc N(log log)/log (2)
x
is means that, to have a 90% chance of eliminating a population of size 10,000, we require a multiplex
number of two; for a population of size 1 million, we require a multiplex number of three; and for a population
of size 10 billion, we require a multiplex number of four. An encouraging property of these predictions is that, as
multiplex number increases linearly, the population size that we can eliminate increases exponentially. A modest
additional increase in multiplex number can also lead to a much higher chance of eliminating a population of
the same size. For instance, a multiplex number of ve is predicted to provide a > 99% chance of eliminating an
An. gambiae population on the scale of the African continent. Important spatial factors are not considered here.
In the multiplexing experiments demonstrating the RGR approach in D. melanogaster, a reduced eciency
was observed at the second target site as compared to the rst, for reasons that are unclear (but that may or may
not be tied to position within the RGR array12). Presumably this is something that could be solved through sub-
sequent engineering eorts; however, mathematical analysis described in Supplementary TextS1 shows that, at
least for the two-gRNA system, a reduced cleavage rate at one site doesn’t signicantly alter the eective resistant
Figure 5. RGR/Cas9-induced mutations at the yellow and white loci. (A) Schematic of the white and yellow
genes showing the gRNA target sites. Exons are shown as blue boxes, the gRNA target site locations are
indicated by black lines, and the gRNA target site sequences (with black letters indicating protospacer sequences
and red letters indicating PAM sequences) are underlined in yellow for white, purple for yellow. (B) Schematic
of the OA-16 construct utilized in generating mutations. e rst and second gRNAs (targeting white and
yellow, respectively) are shown in grey. Each gRNA has a hammerhead ribozyme 5 (shown in blue) and an
HDV ribozyme 3 (shown in green). e gRNAs are driven by a single Drosophila ubiquitin polII promoter. (C)
Crossing scheme used to generate mutants, and obtained results. Individual male and female ies homozygous
for the OA-16 construct were crossed to individual female and male ies, respectively, of a homozygous vasa-
Cas9 line. Progeny were scored for eye and body color. Percentages correspond to number of ies out of total
cross progeny (+/SEM) exhibiting a mutation for each gRNA. (D) e white and yellow tissues of an OA-16
homozygous male y with no exposure to Cas9 are un-mutated (le), while the white and yellow tissues of a
y generated by crossing OA-16 homozygotes to vasa-Cas9 homozygotes show mosaic expression (right). (E)
Examples of sequences of CRISPR/Cas-induced mutations in white (top) and yellow (bottom). e rst line in
each alignment represents wild-type sequence, and subsequent lines show individual mutant clones.
www.nature.com/scientificreports/
8
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
allele generation rate overall. is is because a reduced cleavage rate at one site also suggests a reduced resistant
allele generation rate at this site, as both H and R alleles are generated through the same cleavage and repair
mechanism. However, this reduction in the resistant allele generation rate is compensated for by the increased
accumulation of partially-resistant composite alleles and their subsequent development into completely resistant
composite alleles. e net eect is that, even if one of two multiplexed gRNAs displays a reduced cleavage rate, the
benets of multiplexing in terms of reduced resistant allele generation are very similar.
Discussion
e possibility of using gene drive systems to suppress and potentially eliminate wild populations has provoked
intense interest over the last decade2, 3, 2022. is excitement has recently been fueled by signicant developments
in genetic engineering, and by the CRISPR revolution, which has enabled scientists to develop homing-based
gene drive systems targeting a range of sites in any genome with relative ease. In terms of population suppres-
sion, recessive lethal and sterility genes are of interest as targets because a homing system targeting these genes
can potentially spread to xation and eliminate the population in the process, even when introduced beginning
with a single drive-containing organism4. While this excitement is warranted, it is highly relevant to determine
how the evolution of homing-resistant alleles could interfere with population suppression drive strategies, and to
determine design criteria to increase stability of the drive for the likely elimination of populations of a given size.
To address this important question, we developed a mathematical model to describe the spread of
a population-suppressing homing allele through a population of An. gambiae, and the impact that a
homing-resistant allele could have on these dynamics. Homing-resistant alleles may originate through several
mechanisms: a) de novo mutations, which occur independently from the drive; b) pre-existing natural variation
in the population, which may be minimized through intelligent selection of homing recognition sites; and c) in
response to the drive, either by errors introduced during HDR or by the cell mending DNA damage via the NHEJ
pathway. While the former two mechanisms are important, we have focused this study on the latter (i.e. resistant
allele formation in response to the drive), as this is expected to occur at a signicantly higher frequency than the
other two mechanisms23. Firstly, de novo mutations occur at a rate of ~109–108 per individual per generation
in species similar to that of interest24, 25 – a rate that is many orders of magnitude smaller than resistant allele
generation in response to the drive. Secondly, the frequency of resistance alleles in the population prior to the
introduction of the drive system may be predicted based on mutation-selection-dri balance26. Unckless et al.23
have shown that pre-existing resistant alleles are likely when the eective population size exceeds 106, a popula-
tion size that would be several orders of magnitude higher for a composite allele. Furthermore, prior screening of
target sites in wild populations could be used to identify pre-existing resistant alleles at frequencies less than 103,
and at frequencies several orders of magnitude smaller for composite alleles. All of this highlights the signicant
relative importance of resistant allele generation in response to the drive.
We discover that, despite promising experimental data reporting extremely high rates of homing in the
germline for recently engineered CRISPR-Cas9-based homing systems6, 8, 9, population suppression is expected
to be moderate and short-lived with current construct architectures due to the generation of homing-resistant
alleles. To minimize chances of a population rebound, current homing rates are adequate; however, reducing the
resistant allele generation rate is crucial. For example, to be 95% sure that resistant alleles will not interfere with
suppression of a population of An. gambiae mosquitoes on the scale of the African continent, the resistant allele
generation rate should be less than ~2 × 1012 per homing event – about nine orders of magnitude smaller than
current resistant allele generation rates9.
While it might be near impossible to achieve resistant allele generation rates this low with a single gRNA
recognizing an exclusive target site, one strategy to mitigate the impact of homing-resistant alleles is to multiplex
gRNAs in the drive system2, 3, 9, 17. By multiplexing gRNAs to target multiple locations within an essential gene,
each site is required to be homing-resistant in order for the composite allele to have the homing-resistant phe-
notype. Our results suggest that the eective resistant allele generation rate becomes exponentially smaller as the
number of multiplexed gRNAs increases, and that a multiplex number of four may be sucient to have a 90%
chance of eliminating an An. gambiae population on the scale of the African continent.
Several approaches to multiplexing gRNAs have been described, including the use of dierent polIII pro-
moters such as U6:1-U6:327, HP128, 29, 7SK28, or tRNA promoters30 to promote expression of individual gRNAs
(Fig.6A,B). While these strategies are eective, they are limited by the fact that most polIII promoters do not
drive temporal and/or tissue-specic expression, which may incur increased tness costs to the organism due
to ubiquitous and continuous gRNA expression. ese strategies also require an individual promoter element
for each gRNA, thereby increasing the overall size of the drive and possibly introducing repetitive elements.
Repetitive DNA sequences have reduced stability31 and previous attempts to build drives with zinc-nger nucle-
ases and TALENs have indeed demonstrated that larger and more repetitive the drive systems are less evolutionar-
ily stable. erefore, it is essential to minimize the size and repetitiveness of the drive32. To circumvent the need to
express each gRNA from a dierent polIII promoter, gRNAs can be anked with self-cleaving ribozymes13, 14, 30, 33
or tRNAs12, 27, 30, 34, which can allow the use of a single temporal and tissue-specic polII promoter to drive expres-
sion of an array of anked gRNAs, thus reducing the overall drive size and repetitiveness of the drive element.
To date, only the tRNAanked gRNA multiplex approach has been validated in a whole animal model12.
Here, we report that the adaptation of another multiplex RGR approach, previously demonstrated in yeast13 and
in mammalian tissue culture cells14, functions eciently in D. melanogaster, at least with respect to the ability to
be processed somatically and induce mutations via NHEJ. We observe highly ecient somatic mutation rates (as
measured by the percentage of individuals possessing mutations) approaching 100% for the rst gRNA and of
60%-86% for the second gRNA, depending on whether Cas9 is either maternally or paternally inherited (Fig.5),
and a high frequency of mutation transmission to the ospring of mutant individuals. While these observa-
tions do not directly indicate that the RGR approach will eectively function to induce HDR in the germline,
www.nature.com/scientificreports/
9
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
they at least indicate that ribozyme-anked gRNAs are processed in vivo and can direct ecient mutagenesis.
Importantly, while we use the same two ribozymes to ank each gRNA (Figs5B and 6C), the RGR approach may
be expanded in the future to incorporate dierent ribozymes than the two tested here35, 36 to reduce repetitiveness,
and the same strategy could be applied to the use of tRNAs other than the tRNAGly (Fig.6D) that was shown to
work by Port et al.12. Furthermore, as recently demonstrated by Yoshioka et al.14 in mammalian cells, the RGR
approach could also be optimized to allow for expression of both the multiplexed gRNAs and CRISPR-Cas9
components from a single polII promoter, further reducing drive size and repetitiveness (Fig.6E). Utilization of
this single promoter-gRNA-CRISPR/Cas9 strategy combined with both the tRNA and RGR approaches may yield
an optimal drive element design, both in terms of eciency and stability (Fig.6F). Finally, it may be important
to utilize improved gRNA backbones to further reduce repetitiveness of the gRNAs37. Overall, the above design
considerations may oer opportunities to engineer compact, evolutionarily stable gene drive cassettes; however,
these ideas are largely untested and the use of multiplexed gRNAs in a functional gene drive system remains to
be demonstrated.
Several assumptions have been made in the modeling portion of this study. Most noteworthy is the description
of an An. gambiae population on the scale of the African continent as randomly mixing. Clearly, the study of gene
drive in An. gambiae at anything beyond the village scale will require an understanding of population structure,
and in fact, even at the village scale there are population considerations regarding gene ow within the An. gam-
biae species complex38, 39. By ignoring population structure, the model described here cannot be used to predict
the spatial pattern or timescale of gene drive40. However, even in the presence of population structure, each target
site on a chromosome represents an opportunity for homing resistance to emerge and, in this sense, we expect
there to be some validity to predictions regarding the resistant allele generation rate required to ensure that a pop-
ulation rebound unlikely (Supplementary Figure3). Recent work by Eckho et al.41 has modeled homing-based
gene drive systems in a spatially-explicit manner with seasonality based on Namawala, Tanzania and the Garki
District, Nigeria. For a seasonal population size of 103–105 mosquitoes, a critical NHEJ rate of 105 was observed,
which is in agreement with our model41. One potential impact of population structure is the emergence of resist-
ance to one guide RNA in one population and to another guide RNA in another population rarely combining due
to low rates of gene ow. However, the suggestion in Supplementary TextS1 that partially resistant composite
Figure 6. Schematic of various proposed strategies for multiplexing gRNAs. (A) A gRNA multiplexing scheme
where the same polIII promoter drives each of two gRNAs, and a polII-driven Cas9 is provided as a separate
transgene. (B) A multiplexing scheme where two dierent polIII promoter drive each of two gRNAs, and a
polII-driven Cas9 is provided as a separate transgene. (C) A multiplexing scheme where each of two gRNAs
are surrounded by a 5 HH ribozyme and a 3 HDV ribozyme, and a polII-driven Cas9 is provided as a separate
transgene. (D) A multiplexing scheme where the two gRNAs are surrounded by copies of the same tRNA (with
a tRNA 5 of the rst gRNA, between gRNAs 1 and 2, and 3 of the second gRNA), and a polII-driven Cas9 is
provided as a separate transgene. (E) A multiplexing scheme where each of two gRNAs are surrounded by a 5
HH ribozyme and a 3 HDV ribozyme, as in (C), but the Cas9 is located on the same transgene, 3 of the gRNAs
and preceded by an IRES. (F) A proposed multiplexing scheme where the rst of two gRNAs is surrounded
by two dierent tRNAs, the second gRNA is anked by the HH and HDV ribozymes (as in (C) and (E)), and
the Cas9 is located on the same transgene, 3 of the gRNAs and preceded by an IRES (as in (E)). Grey triangles
represent polIII promoters; blue triangles are polII promoters; terminators (T) and polyA signals (pA) are
shown in grey ovals; Cas9 is represented as a green rectangle; internal ribosomal entry sequences (IRES) are
grey sequences; gRNA scaolds are shown as grey lines, with red and purple connecting lines representing two
dierent gRNAs; the hammerhead (HH) and HDV ribozymes are shown as blue and green lines, respectively;
and two dierent tRNAs (tRNAGly and a non-specic tRNA) are shown as pink and brown lines, respectively.
www.nature.com/scientificreports/
10
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
alleles are rare due to their low rate of emergence and quick conversion to functional homing alleles would argue
against this concern.
At the molecular level, several assumptions have been made regarding the dynamics of multiplexed gRNAs.
ese have been modeled as independently-acting homing systems; however, the use of multiplexed gRNAs in
a functional gene drive system has yet to be demonstrated and hence these dynamics will be elucidated in future
drive experiments. Potential problems may arise from sequence repetitiveness in the drive element if identical
gRNA backbones and promoters are used (Fig.6), creating the possibility of recombination between identical
sequences12 and thus reducing the overall evolutionary stability of the system42. Furthermore, it is not clear how
the cleavage and homing rates will vary as multiplex number is increased substantially. If the gRNA target sites are
far away from each other (e.g., >1–5 kb), then it is theoretically possible that multiplexing may not be an eective
strategy as homology arms may not be able to eectively direct HDR (Supplementary Figure4A–D), while if they
are close together (e.g., <1 kb), multiplexing may increase homing eectiveness (Supplementary Figure4E–H),
as homology arms may able to eectively direct HDR, although this hypothesis remains to be demonstrated.
Interestingly, a reduction in the homing rate associated with one of the gRNAs may not interfere with our design
criteria for population suppression; however, it is important that we have a good quantitative understanding of the
underlying molecular dynamics of multiplexed gRNAs to make accurate model predictions.
In conclusion, multiplexing gRNAs appears to be a highly eective strategy by which to reduce the eective
resistant allele generation rate and hence to enable the elimination of large populations. Due to the exponential
decrease in resistant allele generation with increasing multiplex number, only a modest number of gRNAs are
needed to achieve population suppression potentially on a continental scale. ese approaches need to be tested
in relevant organisms to accurately describe their dynamics and to conrm their utility for population suppres-
sion strategies. Future studies should address additional sources of resistant alleles, such as de novo mutations and
naturally-occurring genetic variation. Additional strategies for overcoming resistance should also be explored, for
instance, engineering successive gene drive systems, each designed to target dierent essential genes, and releas-
ing these one aer the other1. While both this approach and gRNA multiplexing may be eective for overcoming
resistance, neither has been demonstrated. Given how quickly this eld is advancing, understanding strategies
such as these should be of high priority so that the full potential of homing-based population suppression drives
can be properly evaluated.
Materials and Methods
Modeling CRISPR-Cas9 population genetics. To characterize the basic dynamics of the autosomal
CRISPR-Cas9-based gene drive system targeting a gene required for female fertility9, we represent the CRISPR
homing construct as a single autosomal allele, “H”, with a corresponding wild-type allele, “h”. We denote hom-
ing-resistant alleles (mutant alleles with no associated tness cost) as “R”. e CRISPR construct creates a bias
among gametes of both parental heterozygotes towards gametes having the CRISPR allele. We consider the case
where the homing rate is the same among Hh males and females and denote this as e. We also consider a resistant
allele generation rate that is identical among Hh males and females and denote this as ρ. e homing rate, e, is the
proportion of h gametes in heterozygotes that become H gametes due to the act of homing, and hence the pro-
portion of H gametes arising from heterozygotes of both sexes is equal to (1 + e)/2. e resistant allele generation
rate, ρ, is the proportion of h gametes that become R gametes due to errors introduced during the DNA breakage
and repair process, and hence the proportion of R gametes arising from heterozygotes of both sexes is equal to
ρ/2. e remaining gametes arising from heterozygotes are wild-type, h. is leads to an increase in the frequency
of the H allele in the population and, since HH females are infertile, there is potential for a population crash to
occur under permissive conditions. However, since the R allele represents resistance to homing and hence resist-
ance to the spread of the H allele, it has a selective advantage following emergence and is expected to reverse the
eects of population suppression. e crosses describing this system are shown in Supplementary Figure1 and
eective resistant allele generation rates for higher multiplex numbers are derived in Supplementary Text S1.
Modeling An. gambiae population dynamics. Using An. gambiae as a case study, we adapt the modeling
framework of Deredec et al.4 to describe the spread of the CRISPR and homing-resistant alleles through a dis-
crete, density-dependent population with time steps of one day. In this model, the mosquito life cycle is divided
into four life stages – egg, larva, pupa and adult (both male and female adults are modeled). The daily,
density-independent mortality rates for the juvenile stages are assumed to be identical and are chosen for consist-
ency with the population growth rate in the absence of density-dependent mortality, while the duration of these
stages dier. Additional density-dependent mortality occurs at the larval stage, and we use a density-dependent
equation of the form,
αα=+FL L() /( )
TL, where L is the number of larvae, TL is the duration of the larval stage,
and α is a parameter determining the strength of density-dependence which is chosen to produce the desired
equilibrium density of adult mosquitoes in the population. e form of the density-dependence equation is taken
from Deredec et al.4, and may be thought of as a discrete version of the logistic growth model. It has the desirable
feature that, while density-independent life parameters determine the population growth rate, a single parameter,
α, determines the equilibrium density of adults (the value of this parameter may be determined from
EquationS55, Supplementary TextS1). Adult males mate throughout their lifetime, while adult females mate only
once, soon aer that they emerge. Fecundity rates dier per genotype, with wild-type females laying β eggs per
day, females heterozygous for the homing allele laying
βs(1 )
eggs per day, HH females being infertile, and
females of all other genotypes laying β eggs per day. Here, s represents the fractional reduction in fertility of
females heterozygous for the homing allele. Initial estimates for these and other parameter values are provided in
Supplementary Table1. Equations describing this system are provided in Supplementary TextS1.
www.nature.com/scientificreports/
11
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
We use a stochastic implementation of t his model to capture the random eects at low population sizes, for instance
when the CRISPR-Cas9 system is causing signicant population suppression. We assume that the number of eggs
produced per day by females follows a Poisson distribution, the number of eggs having each genotype follows a multi-
nomial distribution, and all survival/death events follow a Bernoulli distribution. Finally, female mate choice follows a
binomial distribution with probabilities given by the relative frequency of each male genotype in the population.
Construct Assembly. Gibson enzymatic assembly (EA) cloning method was used for all cloning43. To gen-
erate plasmid OA-16, components were cloned into the multiple cloning site (MCS) of a commonly used plasmid
in the lab for D. melanogaster transformation that contains the white gene as a marker and an attB-docking site.
Specically, the Drosophila ubiquitin promoter19 was amplied from D. melanogaster genomic DNA using prim-
ers OA16-1 and OA16-2, and the SV40 3UTR fragment was amplied from template pMos-3xP3-DsRed-attp
(addgene plasmid #52904) using primers OA16-3 and OA16-4. e two RGRs were generated via sequential
PCRs using primers OA16-5 and OA16-6 for the rst PCR and OA16-5 and OA16-7 for the white RGR, and
primers OA16-8 and OA16-5 for the rst PCR and OA16-9 and OA1610 for the yellow RGR. e construct was
assembled in one step: the D. melanogaster attB stock plasmid was digested with AscI and XbaI, and the ubiquitin
promoter, white RGR, yellow RGR, and the SV40 3UTR were cloned in via EA cloning. A list of primer sequences
used in the above construct assembly can be found in Supplementary Table2.
Fly Culture and Strains. Fly husbandry and crosses were performed under standard conditions at 25 °C. Rainbow
Transgenics (Camarillo, CA) carried out all of the y injections. e OA-16 construct was integrated into Bloomington
Stock Center (BSC) y strain 86Fa (BSC #24485: y1 M{vas-int.Dm}ZH-2A w*; M{3xP3-RFP.attP’}ZH-68E), and y
stock BSC#51324 (w[1118]; PBac{y[+mDint2] = vas-Cas9}VK00027) was used as the source of vasa-Cas9. For balanc-
ing chromosomes, y stocks BSC#39631 (w[*]; wg[Sp-1]/CyO; P{r y[+t7.2] = neoFRT}82B lsn[SS6]/TM6C, Sb[1]) and
BSC#2555 (CyO/sna[Sco]) were used. Homozygous stocks were rst generated for 86Fa-OA-16 ies via use of balancer
ies. en, single homozygous female virgins and males were crossed out in triplicate to single male and female vir-
gins, respectively, from the vasa-Cas9 line. e ospring (1765 in total) were scored for body color and eye color. e
standard error of the mean (SEM) was calculated for each cross type and each phenotype using standard procedures.
Sequencing to conrm mutations. Genomic DNA was extracted from single mutant ies using Qiagen
DNeasy Blood and Tissue Kit (Hilden, Germany), and PCRs were set up using standard protocols to amplify
regions of the white (primer set OA16-S1/OA16-S2) and yellow (primer set OA16-S3/OA16-S4) genes that span
the cleavage site. Sequencing was performed by Source Bioscience (Nottingham, UK). Primer sequences can be
found in Supplementary Table2.
References
1. Burt, A. Site-specic selsh genes as tools for the control and genetic engineering of natural populations. Proc. Biol. Sci. 270,
921–928, doi:10.1098/rspb.2002.2319 (2003).
2. Champer, J., Buchman, A. & Abari, O. S. Cheating evolution: engineering gene drives to manipulate the fate of wild populations.
Nat. ev. Genet. 17, 146–159, doi:10.1038/nrg.2015.34 (2016).
3. Esvelt, . M., Smidler, A. L., Catteruccia, F. & Church, G. M. Concerning NA-guided gene drives for the alteration of wild
populations. Elife 3, e03401, doi:10.7554/eLife.03401 (2014).
4. Deredec, A., Burt, A. & Godfray, H. C. J. The population genetics of using homing endonuclease genes in vector and pest
management. Genetics 179, 2013–2026, doi:10.1534/genetics.108.089037 (2008).
5. Deredec, A., Godfray, H. C. J. & Burt, A. equirements for eective malaria control with homing endonuclease genes. Proc. Natl.
Acad. Sci. USA 108, E874–80, doi:10.1073/pnas.1110717108 (2011).
6. Gantz, V. M. & Bier, E. Genome editing. e mutagenic chain reaction: a method for converting heterozygous to homozygous
mutations. Science 348, 442–444, doi:10.1126/science.aaa5945 (2015).
7. DiCarlo, J. E., Chavez, A., Dietz, S. L., Esvelt, . M. & Church, G. M. Safeguarding CISP-Cas9 gene drives in yeast. Nat. Biotechnol.
33, 1250–1255, doi:10.1038/nbt.3412 (2015).
8. Gantz, V. M. et al. Highly ecient Cas9-mediated gene drive for population modication of the malaria vector mosquito Anopheles
stephensi. Proc. Natl. Acad. Sci. USA 112, E6736–43, doi:10.1073/pnas.1521077112 (2015).
9. Hammond, A. et al. A CISP-Cas9 gene drive system targeting female reproduction in the malaria mosquito vector Anopheles
gambiae. Nat. Biotechnol. 34, 78–83, doi:10.1038/nbt.3439 (2016).
10. Alphey, L. Genetic control of mosquitoes. N. Biotechnol (2016).
11. Noble, C., Olejarz, J., Esvelt, ., Church, G. & Nowa, M. Evolutionary dynamics of CISP gene drives. Sci. Advances doi:10.1126/
sciadv.1601964(2017).
12. Port, F. & Bulloc, S. L. Augmenting CISP applications in Drosophila with tNA-aned sgNAs. Nat. Methods doi:10.1038/
nmeth.3972 (2016).
13. Gao, Y. & Zhao, Y. Self-processing of ribozyme-aned NAs into guide NAs in vitro and in vivo for CISP-mediated genome
editing. J. Integr. Plant Biol. 56, 343–349, doi:10.1111/jipb.v56.4 (2014).
14. Yoshioa, S., Fujii, W., Ogawa, T., Sugiura, . & Naito, . Development of a mono-promoter-driven CISP/Cas9 system in
mammalian cells. Sci. ep. 5, 18341, doi:10.1038/srep18341 (2015).
15. Bruce-Chwatt, L. J. e Gari Project. esearch on the epidemiology and control of malaria in the Sudan savanna of West Africa.
Trans.  . Soc. Trop. Med. Hyg. 75, 190–191, doi:10.1016/0035-9203(81)90085-7 (1981).
16. Taen, W. & Scott, T. W. Ecological Aspects for Application of Genetically Modied Mosquitoes. (Springer Science & Business Media,
2003).
17. Galizi, . et al. A CISP-Cas9 sex-ratio distortion system for genetic control. Sci. ep. 6, 31139, doi:10.1038/srep31139 (2016).
18. Bassett, A. ., Tibbit, C., Ponting, C. P. & Liu, J.-L. Highly ecient targeted mutagenesis of Drosophila with the CISP/Cas9
system. Cell ep. 4, 220–228, doi:10.1016/j.celrep.2013.06.020 (2013).
19. Abari, O. S., Oliver, D., Eyer, . & Pai, C.-Y. An Entry/Gateway® cloning system for general expression of genes with molecular tags
in Drosophila melanogaster. BMC Cell Biol. 10, 1–11, doi:10.1186/1471-2121-10-8 (2009).
20. Sinins, S. P. & Gould, F. Gene drive systems for insect disease vectors. Nat. ev. Genet. 7, 427–435, doi:10.1038/nrg1870 (2006).
21. Burt, A. Heritable strategies for controlling insect vectors of disease. Philos. Trans. . Soc. Lond. B Biol. Sci 369, 20130432–20130432,
doi:10.1098/rstb.2013.0432 (2014).
www.nature.com/scientificreports/
12
Scientific RepoRts | 7: 3776 | DOI:10.1038/s41598-017-02744-7
22. Gould, F. & Fred, G. Broadening the Application of Evolutionarily Based Genetic Pest Management. Evolution 62, 500–510,
doi:10.1111/j.1558-5646.2007.00298.x (2008).
23. Uncless, . L., Clar, A. G. & Messer, P. W. Evolution of resistance against CISP/Cas9 gene drive. bioxiv 058438,
doi:10.1101/058438 (2016).
24. Haag-Liautard, C. et al. Direct estimation of per nucleotide and genomic deleterious mutation rates in Drosophila. Nature 445,
82–85, doi:10.1038/nature05388 (2007).
25. eightley, P. D. et al. Estimation of the Spontaneous Mutation ate in Heliconius melpomene. Mol. Biol. Evol. 32, 239–243,
doi:10.1093/molbev/msu302 (2014).
26. Hermisson, J. So Sweeps: Molecular Population Genetics of Adaptation From Standing Genetic Variation. Genetics 169, 2335–2352,
doi:10.1534/genetics.104.036947 (2005).
27. Port, F., Chen, H.-M., Lee, T. & Bulloc, S. L. Optimized CISP/Cas tools for ecient germline and somatic genome engineering
in Drosophila. Proc. Natl. Acad. Sci. USA 111, E2967–76, doi:10.1073/pnas.1405500111 (2014).
28. abadi, A. M., Ousterout, D. G., Hilton, I. B. & Gersbach, C. A. Multiplex CISP/Cas9-based genome engineering from a single
lentiviral vector. Nucleic Acids es. 42, e147–e147, doi:10.1093/nar/gu749 (2014).
29. anganathan, V., Wahlin, ., Maruotti, J. & Zac, D. J. Expansion of the CISP–Cas9 genome targeting space through the use of
H1 promoter-expressed guide NAs. Nat. Commun. 5, 4516, doi:10.1038/ncomms5516 (2014).
30. Schwartz, C. M., Hussain, M. S., Blenner, M. & Wheeldon, I. Synthetic NA Polymerase III Promoters Facilitate High-Eciency
CISP–Cas9-Mediated Genome Editing in Yarrowia lipolytica. ACS Synth. Biol. 5, 356–359, doi:10.1021/acssynbio.5b00162 (2016).
31. Bzyme, M. & Lovett, S. T. Instability of repetitive DNA s equences: the role of replication in multiple mechanisms. Proc. Natl. Acad.
Sci. USA 98, 8319–8325, doi:10.1073/pnas.111008398 (2001).
32. Simoni, A. et al. Development of synthetic selsh elements based on modular nucleases in Drosophila melanogaster. Nucleic Acids
es. 42, 7461–7472, doi:10.1093/nar/gu387 (2014).
33. Nissim, L., Perli, S. D., Fr idin, A., Perez-Pinera, P. & Lu, T. . Multiplexed and programmable regulation of gene networs with an
integrated NA and CISP/Cas toolit in human cells. Mol. Cell 54, 698–710, doi:10.1016/j.molcel.2014.04.022 (2014).
34. Xie, ., Minenberg, B. & Yang, Y. Boosting CISP/Cas9 multiplex editing capability with the endogenous tNA-processing
system. Proc. Natl. Acad. Sci. USA 112, 3570–3575, doi:10.1073/pnas.1420294112 (2015).
35. oth, A. et al. A widespread self-cleaving ribozyme class is revealed by bioinformatics. Nat. Chem. Biol. 10, 56–60, doi:10.1038/
nchembio.1386 (2014).
36. Wilson, T. J. & Lilley, D. M. J. Do the hairpin and VS ribozymes share a common catalytic mechanism based on general acid-base
catalysis? A critical assessment of available experimental data. NA 17, 213–221, doi:10.1261/rna.2473711 (2011).
37. yme, S. B., Ahmetova, L., Montague, T. G., Valen, E. & Schier, A. F. Internal guide NA interactions interfere with Cas9-mediated
cleavage. Nat. Commun. 7, 11750, doi:10.1038/ncomms11750 (2016).
38. Taylor, C. et al. Gene ow among populations of the malaria vector, Anopheles gambiae, in Mali, West Africa. Genetics 157, 743–750
(2001).
39. Tripet, F., Dolo, G. & Lanzaro, G. C. Multilevel analyses of genetic dierentiation in Anopheles gambiae s.s. reveal patterns of gene
ow important for malaria-ghting mosquito projects. Genetics 169, 313–324, doi:10.1534/genetics.104.026534 (2005).
40. North, A., Burt, A., Godfray, H. C. J. & Bucley, Y. Modelling the spatial spread of a homing endonuclease gene in a mosquito
population. J. Appl. Ecol. 50, 1216–1225, doi:10.1111/1365-2664.12133 (2013).
41. Echo, P. A., Wenger, E. A., G odfray, H. C. J. & Burt, A. Impact of mosquito gene drive on malaria elimination in a computational
model with explicit spatial and temporal dynamics. Proc. Natl. Acad. Sci. USA 114, E255–E264, doi:10.1073/pnas.1611064114
(2017).
42. Jac, B. . et al. Predicting the Genetic Stability of Engineered DNA Sequences with the EFM Calculator. ACS Synth. Biol 4, 939–943,
doi:10.1021/acssynbio.5b00068 (2015).
43. Gibson, D. G. et al. Enzymatic assembly of DNA molecules up to several hundred ilobases. Nat. Methods 6, 343–345, doi:10.1038/
nmeth.1318 (2009).
Acknowledgements
This work was supported by generous University of California, Riverside (UCR) laboratory start-up funds
awarded to O.S.A., a private donation by www.MaxMind.com awarded to O.S.A., a US National Institutes of
Health (NIH) K22 grant (5K22AI113060-02) and an NIH-R21 grant (R21AI123937-01) awarded to O.S.A., a
UC MEXUS grant (CN-15-47) awarded to J.M.M., and a gi from e Parker Foundation to the University of
California, San Francisco, Global Health Group Malaria Elimination Initiative.
Author Contributions
J.M.M. and H.M.S. developed mathematical models with input from O.S.A. O.S.A. and A.B. developed and tested
the ribozyme multiplexing technique in Drosophila melanogaster. All authors contributed to the writing of the
manuscript.
Additional Information
Supplementary information accompanies this paper at doi:10.1038/s41598-017-02744-7
Competing Interests: e authors declare that they have no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© e Author(s) 2017
... If the drive construct only has significant fitness effects as a homozygote, it can rise to high frequencies; in particular, if the target gene is necessary for female fertility, this provides an effective way to suppress the population (3,4). An important question is how resilient gene-drive systems are to the evolution of resistance (1,(5)(6)(7), which could arise via singlenucleotide mutations or single-nucleotide polymorphisms (SNPs) or by an imperfect endjoining repair process, called nonhomologous end joining (NHEJ) (3,8). Although in caged populations, resistance to suppression has been observed via indels induced at a nonconserved target site within a gene critical for female fertility (3), drive systems where the chosen target site has high conservation have been found to be resilient (4). ...
... Although empirically, we understand well the rate at which these processes occur in individuals, from characterization of the mutation rate (μ) and the net NHEJ rate (ν), both per generation, it is also crucially important to know 1) the population size and 2) the fraction of potential resistance mutants that are actually functional. Rather than the per-individual rate, the population-level mutation rate is critical, as it measures how likely such mutations are to arise in the population (5)(6)(7). For example, if a resistance mutant is produced per individual at rate 10 −7 and the population size is N = 10 4 , the population-level mutation rate is 10 −3 , and it will typically take 1,000 generations before a single resistance mutant is generated in a population, and so resistance is unlikely to arise, given that drive typically acts to suppress a population on a timescale of much less than 100 generations; on the other hand, if N = 10 8 , 10 such mutants are generated every generation, and so resistance mutants with a selective advantage compared to drive are very likely to arise and fix before population elimination. ...
... A promising antiresistance strategy is to exploit multiple redundant target sites for cleavage by the nuclease (1,7,10,11). If there are multiple guide RNAs (gRNAs) or target cut sites within a single locus, since a single successful cut, followed by homology directed repair, is sufficient for the drive to be copied, all sites must develop resistance mutants in order for the non-Mendelian transmission of drive to fail. ...
Article
Full-text available
Evolution of resistance is a major barrier to successful deployment of gene-drive systems to suppress natural populations, which could greatly reduce the burden of many vector-borne diseases. Multiplexed guide RNAs (gRNAs) that require resistance mutations in all target cut sites are a promising antiresistance strategy since, in principle, resistance would only arise in unrealistically large populations. Using stochastic simulations that accurately model evolution at very large population sizes, we explore the probability of resistance due to three important mechanisms: 1) nonhomologous end-joining mutations, 2) single-nucleotide mutants arising de novo, or 3) single-nucleotide polymorphisms preexisting as standing variation. Our results explore the relative importance of these mechanisms and highlight a complexity of the mutation–selection–drift balance between haplotypes with complete resistance and those with an incomplete number of resistant alleles. We find that this leads to a phenomenon where weakly deleterious naturally occurring variants greatly amplify the probability of multisite resistance compared to de novo mutation. This key result provides design criterion for antiresistance multiplexed systems, which, in general, will need a larger number of gRNAs compared to de novo expectations. This theory may have wider application to the evolution of resistance or evolutionary rescue when multiple changes are required before selection can act.
... However, the prevention of resistance to synthetic gene drive elements cannot be fully attained but can be minimized by the vigilant drive design. In their effort to better understand and combat drive resistance in homing-based population suppression systems, Marshall et al. [74] have established a mathematical model to estimate the maximum acceptable resistant allele generation rates to attain stable and long-term suppression of various populations. Findings from this study revealed an inverse correlation between the rate of resistant allele generation and the number of multiplexed gRNAs. ...
... A key challenge preventing the spread of gene drives and cutting back their benefits is the development of resistance. So far, three strategies have been published and shown a positive impact on reducing resistance allele formation in gene drives: (a) the multiplex gRNA expressing system [74,75], (b) the use of the nanos promoter for Cas9 expression [75], and (c) the combination of different gene drives (e.g., homing endonuclease drive with cleave and rescue drive), showing a significant potential to overcome the accumulation of drive-resistant alleles [77,81]. On the contrary, several containment strategies have been successfully proposed and established to mitigate the unintended negative impacts of gene drives on the environment. ...
Article
Full-text available
Ongoing pest and disease outbreaks pose a serious threat to human, crop, and animal lives, emphasizing the need for constant genetic discoveries that could serve as mitigation strategies. Gene drives are genetic engineering approaches discovered decades ago that may allow quick, super-Mendelian dissemination of genetic modifications in wild populations, offering hopes for medicine, agriculture, and ecology in combating diseases. Following its first discovery, several naturally occurring selfish genetic elements were identified and several gene drive mechanisms that could attain relatively high threshold population replacement have been proposed. This review provides a comprehensive overview of the recent advances in gene drive research with a particular emphasis on CRISPR-Cas gene drives, the technology that has revolutionized the process of genome engineering. Herein, we discuss the benefits and caveats of this technology and place it within the context of natural gene drives discovered to date and various synthetic drives engineered. Later, we elaborate on the strategies for designing synthetic drive systems to address resistance issues and prevent them from altering the entire wild populations. Lastly, we highlight the major applications of synthetic CRISPR-based gene drives in different living organisms, including plants, animals, and microorganisms.
... The DCDs that we model are the direct DCD ( A E/ B N A / C G B ), and indirect DCD ( A G A / B N/ C G B ), each composed of three elements (downstream to upstream: A, B, C). In each case, we have contrasted our results with the behaviour of a self-perpetuating drive ( A N A ) for which more experimental and other computational results have been reported on the effect of fitness costs [20,46], resistance alleles [20,46,47], cut-rate [20], and deposition [46,48]. Despite using the same underlying inheritance biasing mechanism, we find that the performance of the DCDs can differ from a single-element self-perpetuating drive in important ways. ...
Article
Full-text available
The introgression of genetic traits through gene drive may serve as a powerful and widely applicable method of biological control. However, for many applications, a self-perpetuating gene drive that can spread beyond the specific target population may be undesirable and preclude use. Daisy-chain gene drives have been proposed as a means of tuning the invasiveness of a gene drive, allowing it to spread efficiently into the target population, but be self-limiting beyond that. Daisy-chain gene drives are made up of multiple independent drive elements, where each element, except one, biases the inheritance of another, forming a chain. Under ideal inheritance biasing conditions, the released drive elements remain linked in the same configuration, generating copies of most of their elements except for the last remaining link in the chain. Through mathematical modelling of populations connected by migration, we have evaluated the effect of resistance alleles, different fitness costs, reduction in the cut-rate, and maternal deposition on two alternative daisy-chain gene drive designs. We find that the self-limiting nature of daisy-chain gene drives makes their spread highly dependent on the efficiency and fidelity of the inheritance biasing mechanism. In particular, reductions in the cut-rate and the formation of non-lethal resistance alleles can cause drive elements to lose their linked configuration. This severely reduces the invasiveness of the drives and allows for phantom cutting, where an upstream drive element cuts a downstream target locus despite the corresponding drive element being absent, creating and biasing the inheritance of additional resistance alleles. This phantom cutting can be mitigated by an alternative indirect daisy-chain design. We further find that while dominant fitness costs and maternal deposition reduce daisy-chain invasiveness, if overcome with an increased release frequency, they can reduce the spread of the drive into a neighbouring population.
... pre-existing sequence variations (whether r1 or r2) or failed homing attempts must have inhibited all target sequences to fully inhibit further drive (19)(20)(21)(22)(23)(24). Such a system is furthermore less likely to result in functional resistant alleles, given the multiple disruptions to the target gene and will function most effectively if non-functional mutations result in some fitness cost, as would likely be expected of most genes. ...
Preprint
Full-text available
Aedes aegypti , the yellow fever mosquito, is the main vector of several major pathogens including yellow fever, dengue, Zika and chikungunya viruses. Classical mosquito control strategies, mainly utilizing insecticides, have had success in controlling other mosquito vectors in recent years, but are much less useful against Ae. aegypti , and even these methods are threatened by rising insecticide resistance. This has stimulated interest in new mosquito control mechanisms, notably genetic systems such as gene drives. However, the development of CRISPR/Cas9 gene drive systems has faced challenges such as low inheritance biasing rate, the emergence of resistance alleles, and the possibility of spreading beyond the intended population. Here, we test the regulatory sequences from the Ae. aegypti benign gonial cell neoplasm ( bgcn ) homolog to express Cas9 in the germline to find an expression timing more conducive to homing. We also created a separate multiplexing (targeting multiple different sites within the target gene) sgRNA-expressing homing cassette inserted into the Ae. aegypti kynurenine 3-monooxygenase ( kmo ) gene to limit the consequences of resistance alleles. This creates a ‘split’ gene drive such that one part does not drive, allowing control over geographic spread and temporal persistence. When combined, these two elements provide highly effective germline cutting at the kmo locus and act as a gene drive. Our target genetic element was driven through a cage trial population such that carrier frequency of the element increased from 50% to up to 89% of the population despite significant fitness costs to kmo insertions. Deep sequencing suggests that the multiplexing design could mitigate resistance allele formation in our gene drive system. Significance statement Mosquito-borne diseases affect millions of people worldwide, with the yellow fever mosquito ( Aedes aegypti ) being the principal vector of many viral diseases. Effective measures for controlling this mosquito are sorely needed. Gene drive systems have arisen as a potential tool for mosquito control due to their ability of biasing inheritance of a trait into a target population. Here, we assess a split gene drive, based on CRISPR/Cas9 endonuclease technology driving a target element into the mosquito population. Evaluated over successive generations in a replicated cage trial, the drive successfully biased its inheritance, increasing in frequency from 50% to up to 89%. Our results are encouraging for the potential use of this type of contained gene drive system for mosquito control in endemic areas.
... Including multiple guide RNAs designed against numerous sequences at the target loci, also known as "multiplexing," is one frequently discussed mitigation against this (Carrami et al., 2018). Preexisting sequence variations or failed homing attempts must inhibit all target sequences simultaneously to inhibit the drive and are therefore less likely to generate functional resistant mutants (Marshall et al., 2017;Champer et al., 2018Champer et al., , 2020bOberhofer et al., 2018;Champer S. E. et al., 2020). However, the small target site of the exon-splice junction in dsx means that multiplexing guide RNAs would be difficult to engineer for this gene and would likely be an issue in using homologous targets in other species. ...
Article
Full-text available
Sex determination pathways in insects are generally characterised by an upstream primary signal, which is highly variable across species, and that regulates the splicing of a suite of downstream but highly-conserved genes (transformer, doublesex and fruitless). In turn, these downstream genes then regulate the expression of sex-specific characteristics in males and females. Identification of sex determination pathways has and continues to be, a critical component of insect population suppression technologies. For example, “first-generation” transgenic technologies such as fsRIDL (Female-Specific Release of Insects carrying Dominant Lethals) enabled efficient selective removal of females from a target population as a significant improvement on the sterile insect technique (SIT). Second-generation technologies such as CRISPR/Cas9 homing gene drives and precision-guided SIT (pgSIT) have used gene editing technologies to manipulate sex determination genes in vivo. The development of future, third-generation control technologies, such as Y-linked drives, (female to male) sex-reversal, or X-shredding, will require additional knowledge of aspects of sexual development, including a deeper understanding of the nature of primary signals and dosage compensation. This review shows how knowledge of sex determination in target pest species is fundamental to all phases of the development of control technologies.
... On the other hand, functional resistant mutations pose a large threat to suppression drives [2,19] and have been found to undergo strong selection in cage trials where the drive is costly [41,68]. Whilst various improvements have been made to suppression drives to reduce the likelihood of resistance arising, including the use of highly constrained gRNA target sites [21] and optimisation of the level and spatiotemporal selectivity of Cas9 expression [22,69], the generation and accumulation of resistant alleles still poses a risk to the long-term success of suppression programmes [70,71]. The fitness cost imposed by zpg D is comparable to (due to its dominance) that of a suppressive gene drive. ...
Article
Full-text available
Gene drives for mosquito population modification are novel tools for malaria control. Strategies to safely test antimalarial effectors in the field are required. Here, we modified the Anopheles gambiae zpg locus to host a CRISPR/Cas9 integral gene drive allele ( zpg D ) and characterized its behaviour and resistance profile. We found that zpg D dominantly sterilizes females but can induce efficient drive at other loci when it itself encounters resistance. We combined zpg D with multiple previously characterized non-autonomous payload drives and found that, as zpg D self-eliminates, it leads to conversion of mosquito cage populations at these loci. Our results demonstrate how self-eliminating drivers could allow safe testing of non-autonomous effector-traits by local population modification. They also suggest that after engendering resistance, gene drives intended for population suppression could nevertheless serve to propagate subsequently released non-autonomous payload genes, allowing modification of vector populations initially targeted for suppression.
Chapter
Molecular techniques like DNA barcoding and genome editing play a significant role in insect pest management. DNA barcoding is employed for identifying, monitoring, dispersal, and establishment of insect pests and natural enemies. Production, formulation, and storage of entomopathogenic fungi can be dramatically improved through biotechnology and genetic engineering. DNA base barcoding enables rapid identification of molecular operational taxonomic unit (MOTU) for the assessment of any taxa. Compared to traditional breeding methods, genome editing technologies provide obvious advantages. The recent establishment of genome editing technique provides the capacity for gene knock-out (KO), knock-in (KI), and/or knock-down (KD) in living cells. The latest genome editing technologies, particularly CRISPR/Cas9, are more efficient and precise to edit genes when the genome sequences for target genes are known. CRISPR/Cas9 technology is probably superseding ZFNs and TALENS. However, the CRIPSR/Cas9 methods are also being improvised, and newer additions have further enhanced functional capabilities with reduced off-target effects. Applications of barcoding, gene editing, and genome engineering in insect pest management are discussed.KeywordsDNA barcodingGenome editingCRISPERTranscription
Chapter
Malaria is a mosquito-borne disease that kills millions of people every year. Existing control tools have been insufficient to eliminate the disease in many endemic regions and additional approaches are needed. Novel vector-control strategies using genetic engineering to create malaria-resistant mosquitoes (population modification) can potentially contribute a new set of tools for mosquito control. Here we review the current mosquito control strategies and the development of transgenic mosquitoes expressing anti-parasite effector genes, highlighting the recent improvements in mosquito genome editing with CRISPR-Cas9 as an efficient and adaptable tool for gene-drive systems to effectively spread these genes into mosquito populations.
Article
Full-text available
Background Gene drives are a genetic engineering method where a suite of genes is inherited at higher than Mendelian rates and has been proposed as a promising new vector control strategy to reinvigorate the fight against malaria in sub-Saharan Africa. Methods Using an agent-based model of malaria transmission with vector genetics, the impacts of releasing population-replacement gene drive mosquitoes on malaria transmission are examined and the population replacement gene drive system parameters required to achieve local elimination within a spatially-resolved, seasonal Sahelian setting are quantified. The performance of two different gene drive systems—“classic” and “integral”—are evaluated. Various transmission regimes (low, moderate, and high—corresponding to annual entomological inoculation rates of 10, 30, and 80 infectious bites per person) and other simultaneous interventions, including deployment of insecticide-treated nets (ITNs) and passive healthcare-seeking, are also simulated. Results Local elimination probabilities decreased with pre-existing population target site resistance frequency, increased with transmission-blocking effectiveness of the introduced antiparasitic gene and drive efficiency, and were context dependent with respect to fitness costs associated with the introduced gene. Of the four parameters, transmission-blocking effectiveness may be the most important to focus on for improvements to future gene drive strains because a single release of classic gene drive mosquitoes is likely to locally eliminate malaria in low to moderate transmission settings only when transmission-blocking effectiveness is very high (above ~ 80–90%). However, simultaneously deploying ITNs and releasing integral rather than classic gene drive mosquitoes significantly boosts elimination probabilities, such that elimination remains highly likely in low to moderate transmission regimes down to transmission-blocking effectiveness values as low as ~ 50% and in high transmission regimes with transmission-blocking effectiveness values above ~ 80–90%. Conclusion A single release of currently achievable population replacement gene drive mosquitoes, in combination with traditional forms of vector control, can likely locally eliminate malaria in low to moderate transmission regimes within the Sahel. In a high transmission regime, higher levels of transmission-blocking effectiveness than are currently available may be required.
Article
Gene drive technology has been proposed to control invasive rodent populations as an alternative to rodenticides. However, this approach has not undergone risk assessment that meets criteria established by Gene Drives on the Horizon, a 2016 report by the National Academies of Sciences, Engineering, and Medicine. To conduct a risk assessment of gene drives, we employed the Bayesian network–relative risk model to calculate the risk of mouse eradication on Southeast Farallon Island using a CRISPR‐Cas9 homing gene drive construct. We modified and implemented the R‐based model “MGDrivE” to simulate and compare 60 management strategies for gene drive rodent management. These scenarios spanned four gene drive mouse release schemes, three gene drive homing rates, three levels of supplemental rodenticide dose, and two timings of rodenticide application relative to gene drive release. Simulation results showed that applying a supplemental rodenticide simultaneously with gene drive mouse deployment resulted in faster eradication of the island mouse population. Gene drive homing rate had the highest influence on the overall probability of successful eradication, as increased gene drive accuracy reduces the likelihood of mice developing resistance to the CRISPR‐Cas9 homing mechanism.
Preprint
Full-text available
The idea of driving genetically modified alleles to fixation in a population has fascinated scientists for over 40 years 1,2 . Potential applications are broad and ambitious, including the eradication of disease vectors, the control of pest species, and the preservation of endangered species from extinction ³ . Until recently, these possibilities have remained largely abstract due to the lack of an effective drive mechanism. CRISPR/Cas9 gene drive (CGD) now promise a highly adaptable approach for driving even deleterious alleles to high population frequency, and this approach was recently shown to be effective in small laboratory populations of insects 4–7 . However, it remains unclear whether CGD will also work in large natural populations in the face of potential resistance mechanisms. Here we show that resistance against CGD will inevitably evolve unless populations are small and repair of CGD-induced cleavage via nonhomologous end joining (NHEJ) can be effectively suppressed, or resistance costs are on par with those of the driver. We specifically calculate the probability that resistance evolves from variants at the target site that are not recognized by the driver's guide RNA, either because they are already present when the driver allele is introduced, arise by de novo mutation, or are created by the driver itself when NHEJ introduces mutations at the target site. Our results shed light on strategies that could facilitate the engineering of a successful drive by lowering resistance potential, as well as strategies that could promote resistance as a possible mechanism for controlling a drive. This study highlights the need for careful modeling of CGD prior to the actual release of a driver construct into the wild.
Article
Full-text available
The alteration of wild populations has been discussed as a solution to a number of humanity’s most pressing ecological and public health concerns. Enabled by the recent revolution in genome editing, clustered regularly interspaced short palindromic repeats (CRISPR) gene drives—selfish genetic elements that can spread through populations even if they confer no advantage to their host organism—are rapidly emerging as the most promising approach. However, before real-world applications are considered, it is imperative to develop a clear understanding of the outcomes of drive release in nature. Toward this aim, we mathematically study the evolutionary dynamics of CRISPR gene drives. We demonstrate that the emergence of drive-resistant alleles presents a major challenge to previously reported constructs, and we show that an alternative design that selects against resistant alleles could greatly improve evolutionary stability. We discuss all results in the context of CRISPR technology and provide insights that inform the engineering of practical gene drive systems.
Article
Full-text available
The renewed effort to eliminate malaria and permanently remove its tremendous burden highlights questions of what combination of tools would be sufficient in various settings and what new tools need to be developed. Gene drive mosquitoes constitute a promising set of tools, with multiple different possible approaches including population replacement with introduced genes limiting malaria transmission, driving-Y chromosomes to collapse a mosquito population, and gene drive disrupting a fertility gene and thereby achieving population suppression or collapse. Each of these approaches has had recent success and advances under laboratory conditions, raising the urgency for understanding how each could be deployed in the real world and the potential impacts of each. New analyses are needed as existing models of gene drive primarily focus on nonseasonal or nonspatial dynamics. We use a mechanistic, spatially explicit, stochastic, individual-based mathematical model to simulate each gene drive approach in a variety of sub-Saharan African settings. Each approach exhibits a broad region of gene construct parameter space with successful elimination of malaria transmission due to the targeted vector species. The introduction of realistic seasonality in vector population dynamics facilitates gene drive success compared with nonseasonal analyses. Spatial simulations illustrate constraints on release timing, frequency, and spatial density in the most challenging settings for construct success. Within its parameter space for success, each gene drive approach provides a tool for malaria elimination unlike anything presently available. Provided potential barriers to success are surmounted, each achieves high efficacy at reducing transmission potential and lower delivery requirements in logistically challenged settings.
Article
Full-text available
Genetic control aims to reduce the ability of insect pest populations to cause harm via the release of modified insects. One strategy is to bias the reproductive sex ratio towards males so that a population decreases in size or is eliminated altogether due to a lack of females. We have shown previously that sex ratio distortion can be generated synthetically in the main human malaria vector Anopheles gambiae, by selectively destroying the X-chromosome during spermatogenesis, through the activity of a naturally-occurring endonuclease that targets a repetitive rDNA sequence highly-conserved in a wide range of organisms. Here we describe a CRISPR-Cas9 sex distortion system that targets ribosomal sequences restricted to the member species of the Anopheles gambiae complex. Expression of Cas9 during spermatogenesis resulted in RNA-guided shredding of the X-chromosome during male meiosis and produced extreme male bias among progeny in the absence of any significant reduction in fertility. The flexibility of CRISPR-Cas9 combined with the availability of genomic data for a range of insects renders this strategy broadly applicable for the species-specific control of any pest or vector species with an XY sex-determination system by targeting sequences exclusive to the female sex chromosome.
Article
Full-text available
The CRISPR/Cas system uses guide RNAs (gRNAs) to direct sequence-specific DNA cleavage. Not every gRNA elicits cleavage and the mechanisms that govern gRNA activity have not been resolved. Low activity could result from either failure to form a functional Cas9-gRNA complex or inability to recognize targets in vivo. Here we show that both phenomena influence Cas9 activity by comparing mutagenesis rates in zebrafish embryos with in vitro cleavage assays. In vivo, our results suggest that genomic factors such as CTCF inhibit mutagenesis. Comparing near-identical gRNA sequences with different in vitro activities reveals that internal gRNA interactions reduce cleavage. Even though gRNAs containing these structures do not yield cleavage-competent complexes, they can compete with active gRNAs for binding to Cas9. These results reveal that both genomic context and internal gRNA interactions can interfere with Cas9-mediated cleavage and illuminate previously uncharacterized features of Cas9-gRNA complex formation.
Article
Full-text available
The oleaginous yeast Yarrowia lipolytica is a valuable microbial host for chemical production because it has a high capacity to synthesize, modify, and store intracellular lipids; however, rapid strain development has been hampered by the limited availability of genome engineering tools. We address this limitation by adapting the CRISPR-Cas9 system from Streptococcus pyogenes for markerless gene disruption and integration in Y. lipolytica. Single gene disruption efficiencies of 92% and higher were achieved when single guide RNAs (sgRNA) were transcribed with synthetic hybrid promoters that combine native RNA polymerase III (Pol III) promoters with tRNA. The Pol III-tRNA hybrid promoters exploit endogenous tRNA processing to produce mature sgRNA for Cas9 targeting. The highest efficiencies were achieved with a SCR1'-tRNA(Gly) promoter and Y. lipolytica codon optimized Cas9 expressed from a UAS1B8-TEF promoter. Cotransformation of the Cas9 and sgRNA expressing plasmid with a homologous recombination donor plasmid resulted in markerless homologous recombination efficiency of over 64%. Homologous recombination was observed in 100% of transformants when nonhomologous end joining was disrupted. The end result of these studies was the development of pCRISPRyl, a modular tool for markerless gene disruption and integration in Y. lipolytica.
Article
CRISPR/Cas9 gene drive (CGD) promises a highly adaptable approach for spreading genetically engineered alleles throughout a species, even if those alleles impair reproductive success. CGD has been shown to be effective in laboratory crosses of insects, yet it remains unclear to what extent potential resistance mechanisms will affect the dynamics of this process in large natural populations. Here we develop a comprehensive population genetic framework for modeling CGD dynamics, which incorporates potential resistance mechanisms as well as random genetic drift. Using this framework, we calculate the probability that resistance against CGD evolves from standing genetic variation, de novo mutation of wildtype alleles, or cleavage-repair by nonhomologous end joining (NHEJ) -- a likely byproduct of CGD itself. We show that resistance to standard CGD approaches should evolve almost inevitably in most natural populations, unless repair of CGD-induced cleavage via NHEJ can be effectively suppressed, or resistance costs are on par with those of the driver. The key factor determining the probability that resistance evolves is the overall rate at which resistance alleles arise at the population level by mutation or NHEJ. By contrast, the conversion efficiency of the driver, its fitness cost, and its introduction frequency have only minor impact. Our results shed light on strategies that could facilitate the engineering of drivers with lower resistance potential, and motivate the possibility to embrace resistance as a possible mechanism for controlling a CGD approach. This study highlights the need for careful modeling of the population dynamics of CGD prior to the actual release of a driver construct into the wild.
Article
We present tRNA-based vectors for producing multiple clustered regularly interspaced short palindromic repeats (CRISPR) single guide RNAs (sgRNAs) from a single RNA polymerase II or III transcript in Drosophila. The system, which is based on liberation of sgRNAs by processing flanking tRNAs, permits highly efficient multiplexing of Cas9-based mutagenesis. We also demonstrate that the tRNA-sgRNA system markedly increases the efficacy of conditional gene disruption by Cas9 and can promote editing by the recently discovered RNA-guided endonuclease Cpf1.
Article
Engineered gene drives - the process of stimulating the biased inheritance of specific genes - have the potential to enable the spread of desirable genes throughout wild populations or to suppress harmful species, and may be particularly useful for the control of vector-borne diseases such as malaria. Although several types of selfish genetic elements exist in nature, few have been successfully engineered in the laboratory thus far. With the discovery of RNA-guided CRISPR-Cas9 (clustered regularly interspaced short palindromic repeats-CRISPR-associated 9) nucleases, which can be utilized to create, streamline and improve synthetic gene drives, this is rapidly changing. Here, we discuss the different types of engineered gene drives and their potential applications, as well as current policies regarding the safety and regulation of gene drives for the manipulation of wild populations.