1
Differential efficacies of Cas nucleases on microsatellites involved in human disorders
and associated off-target mutations
Lucie Poggi
1,2,3
, Lisa Emmenegger
1,4
, Stéphane Descorps-Declère
1,5
, Bruno Dumas
3
, Guy-
Franck Richard
1,2
1 Institut Pasteur, CNRS, UMR3525, 25 rue du Dr Roux, F-75015 Paris, France
2 Sorbonne Université, Collège Doctoral, 4 Place Jussieu, F-75005 Paris, France
3 Biologics Research, Sanofi R&D, 13 Quai Jules Guesde, 94403 Vitry sur Seine, France
4 Present address: Berlin Institute for Medical Systems Biology, Max-Delbrück-Center for
Molecular Medicine in the Helmholtz Association, Robert-Rössle-Strasse 10, 13125 Berlin,
Germany.
5 Institut Pasteur, Bioinformatics and Biostatistics Hub, Department of Computational
Biology, USR3756 CNRS, F-75015 Paris, France
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
2
Abstract
Microsatellite expansions are the cause of more than 20 neurological or developmental human
disorders. Shortening expanded repeats using specific DNA endonucleases may be envisioned
as a gene editing approach. Here, a new assay was developed to test several CRISPR-Cas
nucleases on microsatellites involved in human diseases, by measuring at the same time
double-strand break rates, DNA end resection and homologous recombination efficacy. Broad
variations in nuclease performances were detected on all repeat tracts.
Streptococcus pyogenes Cas9 was the most efficient of all. All repeat tracts did inhibit
double-strand break resection. We demonstrate that secondary structure formation on the
guide RNA was a major determinant of nuclease efficacy. Using deep sequencing, off-target
mutations were assessed genomewide. Out of 221 CAG/CTG or GAA/TTC trinucleotide
repeats of the yeast genome, three were identified as carrying statistically significant low
frequency mutations, corresponding to off-target effects.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
3
Introduction
A growing number of neurological disorders were identified to be linked to microsatellite
expansions (Orr and Zoghbi, 2007). Each disease is associated to a repeat expansion at a
specific locus (Table 1). No cure exists for any of these dramatic disorders. Shortening the
expanded array to non-pathological length could suppress symptoms of the pathology and
could be used as a new gene therapy approach (Richard, 2015). Indeed, when a trinucleotide
repeat contraction occurred during transmission from father to daughter of an expanded
myotonic dystrophy type 1 allele, clinical examination of the daughter showed no sign of the
disease (O’Hoy et al., 1993) (Shelbourne et al., 1992).
In order to induce a double-strand break (DSB) into a microsatellite, different types of
nucleases can be used: meganucleases, Zinc Finger Nucleases (ZFN), Transcription activator-
like effector nucleases (TALEN) and CRISPR-Cas9. Previous experiments using the I-SceI
meganuclease to induce a DSB into a CTG repeat tract showed that repair occurred by
annealing between the flanking CTG repeats (Richard et al., 1999). Later on, ZFNs were used
to induce DSBs into CAG or CTG repeats, which mostly led to contractions in CHO cells
(Mittelman et al., 2009) and in a HEK293 cell GFP reporter assay (Santillan et al., 2014). As
only one arm was enough to induce a DSB into the repeat tract and since CAG zinc fingers
can recognize CTG triplets and vice versa, the authors concluded that the specificity was too
low for further medical applications. As a proof of concept of the approach, a myotonic
dystrophy type 1 CTG repeat expansion was integrated into a yeast strain. A TALEN was
designed to recognize and cut the CTG triplet repeat and was very efficient at shortening it in
yeast cells (>99% cells showed contraction) and highly specific as no other mutation was
detected (Richard et al., 2014). The TALEN was shown to induce specific repeat contractions
through single-strand annealing (SSA) by a RAD52, RAD50 and SAE2 dependent mechanism
(Mosbach et al., 2018). As a proof of concept of the approach, a myotonic dystrophy type 1
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
4
CTG repeat expansion was integrated into a yeast strain. A TALEN was designed to
recognize and cut the CTG triplet repeat and was very efficient at shortening it in yeast cells
(>99% cells showed contraction) and highly specific as no other mutation was detected
(Richard et al., 2014).
The CRISPR-Cas system is the easiest to manipulate and to target any locus, as sequence
recognition is based on the complementarity to a guide RNA (gRNA). To recognize its
sequence, Cas9 requires a specific protospacer adjacent motif (PAM) that varies depending on
the bacterial species of the Cas9 gene. The most widely used Cas9 is wild-type Streptococcus
pyogenes Cas9 (SpCas9) (Cong et al., 2013). Its Protospacer Adjacent Motif (PAM) is NGG
and induces a blunt cut 3-4 nucleotides away from it, through concerted activation of two
catalytical domains, RuvC and HNH, each catalyzing one single-strand break (SSB). Issues
were recently raised about the specificity of SpCas9, leading to the engineering of more
specific variants. In eSpCas9, three positively charged residues interacting with the phosphate
backbone of the non-target strand were neutralized, conferring an increased specificity
(Kleinstiver et al., 2016). Similarly, Cas9-HF1 was mutated on 4 residues interacting through
hydrogen bonds with the target strand (Slaymaker et al., 2016). Staphylococcus aureus is a
smaller Cas9, its PAM is NNGRRT, having a similar structure to SpCas9 with two catalytic
sites. Finally, type V CRISPR-Cas, Cpf1 nucleases, exhibit very different features including a
T-rich PAM located 3’ of target DNA and making staggered cuts leaving five-nucleotide
overhangs by iterative activation of a single RuvC catalytic site (Zetsche et al., 2015).
Cutting repeated sequences like microsatellites may be difficult due to stable secondary
structures that may form either on target DNA or on the guide RNA, making some repeats
more or less permissive to nuclease recognition and cleavage. In addition, secondary structure
formation could impede DSB resection or later repair steps. Eukaryotic genomes contain
thousands of identical microsatellites, therefore the specificity issue may become a real
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
5
problem when targeting one single locus. Here we developed an in vivo assay in the yeast
Saccharomyces cerevisiae in order to test different nucleases belonging to the CRISPR-Cas
family on synthetic microsatellites associated to human disorders. Our experiments revealed
that these sequences may be cut, with surprisingly different efficacies between nucleases and
between microsatellites. SpCas9 was the most efficient and nuclease efficacy relied mainly on
gRNA stability, strongly suggesting that secondary structures are the limiting factor in
inducing a DSB in vivo. DSB resection was decreased to different levels in all repeated tracts.
In addition, we analyzed off-target mutations genomewide and found that three microsatellites
with similar sequences were also edited by the nuclease. The mutation pattern was different
depending on the microsatellite targeted.
Materials and Methods
Yeast plasmids
A synthetic cassette (synYEGFP) was ordered from ThermoFisher (GeneArt). It is a pUC57
vector containing upstream and downstream CAN1 homology sequences flanking a bipartite
eGFP gene interrupted by the I-Sce I recognition sequence (18 bp) under the control of the
TEF1 promoter and followed by the CYC1 terminator. The TRP1 selection marker along with
its own promoter and terminator regions was added downstream the eGFP sequences (Figure
1A). The I-Sce I site was flanked by Sap I recognition sequences, in order to clone the
different repeat tracts. Nine out the 10 repeat tracts were ordered from ThermoFisher
(GeneArt) as 151 bp DNA fragments containing 100 bp of repeated sequence flanked by Sap
I sites. The last repeat (GGGGCC) was ordered from Proteogenix. All these repeat tracts were
cloned at the Sap I site of synYEGFP by standard procedures, to give plasmids pLPX101 to
pLPX110 (Supplemental Table S1). All nucleases were cloned in a centromeric yeast plasmid
derived from pRS415 (Sikorski and Hieter, 1989), carrying a LEU2 selection marker. Each
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which wasthis version posted November 29, 2019. ; https://doi.org/10.1101/857714doi: bioRxiv preprint
