1 Introduction
A highly accurate, cost-effective and easy-to-use technology, the CRISPR-Cas genome editing system has
been favoring the development of promising innovations [25]. Among them, CRISPR-Cas9 gene drive
[1], which aims to spread a trait of interest in a wild type population in a relatively short number of
generations [26]. Application fields are numerous, and include i) the eradication of insect-borne diseases
[10, 18, 26]; ii) the elimination of herbicide and pesticide resistance in pest populations [31]; iii) the
control of destructive invasive species [19, 21]; iv) the conservation of biodiversity by spreading beneficial
traits in endangered species [17, 35].
Targeting sexually reproducing species, CRISPR-Cas9 gene drive biases the transmission of an al-
lele from a parent to its offspring. This biased inheritance occurs through gene conversion (also called
“homing” [15]): in a heterozygous cell, the gene drive cassette present on one chromosome induces a
double-strand break at a specific target site on the homologous chromosome, and the repair process du-
plicates the cassette. Overall, this process increases the chances of transmitting the gene drive cassette
compared to its wild-part counterpart, and the mechanism repeats through the generations. Gene con-
version can potentially take place at different timings of the life cycle: from very early on, in the zygote,
meaning that potentially every single cell of the individual could become homozygous for the gene drive,
to, in the germline, where only the gametes are converted.
Gene drives can be classified into two main categories depending on the purpose of their use [16, 20].
A “replacement drive” is aimed at spreading a genetic modification in order to introduce an important and
durable feature in the natural population. Population size is then not significantly affected and the drive
construct may in principle persist indefinitely in the environment. A “suppression drive” on the other
hand is meant to reduce population size by spreading a detrimental trait, such as a sex ratio distorter
[29] or by altering fertility [26], for example. The term “eradication drive” can be used for the extreme
case where population extinction is the aim.
As with any new tool, it is essential to balance risks (safety) and benefits (efficacy) of the technique
before running any field trials. Experiments currently conducted in laboratories provide small- to medium-
scale information; mathematical models can help to extend these empirical results and identify the features
that are the most important in determining the dynamics at larger scales [13].
Early gene drive models [11, 15, 40] used classical population genetics frameworks, and considered
discrete non-overlapping generations in a well-mixed population. These simplifications helped to draw
general conclusions, but it is important to challenge them. First of all, most of the species targeted
in the context of gene drive do not have synchronous generations (for instance mosquitoes [18, 23, 10,
26], flies [19], mice [21]). Secondly, the assumption of a single well-mixed collection of individuals living
across a uniform space is usually not realistic. In fact, most of the natural landscapes are heterogeneous.
Individuals are also more likely to interact with others that are in closer proximity, which might result in
local genetic variations. Finally, releases of transgenic individuals are limited in range, which is another
factor of spatial heterogeneity.
Taking into account spatio-temporal dynamics of the population size is another key step towards
more realistic models. For the sake of simplicity, most early models focused on allele frequencies and
considered a constant population density. However in the context of gene drive, the introduction of
maladapted transgenic individuals can lead to the reduction (or even extinction) of the population [16].
When considering a spatially structured population, variations in population density naturally generate a
demographic flux from denser to less dense areas. This demographic flux is directed in opposition to the
spread of the drive allele. It was previously shown [20] that the advantage conferred by gene conversion
may nevertheless counteract the demographic effect linked to the fitness cost.
The main goal of this paper is to clarify the impact of variations in population density over the course
of drive propagation over space.
We study partial differential equations which follow the propagation of the drive in space and time.
We explore numerically and analytically two models: a first model based on perfect conversion in the
zygote, already introduced in [20] in a spatially structured population, corresponding to an idealized
case where gene conversion always succeeds; second, a more realistic model with partial conversion and
presence of heterozygous individuals, already studied in [35] in a well-mixed, non spatial population. In
order to investigate the possible spreading of gene drives through space after local introduction, we focus
on the description of traveling waves solutions, that is, particular solutions which are stationary in a
frame moving at constant speed. Our analysis goes beyond [20] by several means: we extend it to the
case of partial conversion, and we systematically analyze the case where the demographic effects are the
4