Replicating human disease in rodents: the good and the bad of CRISPR/Cas9 genome editing
Guillaume Pavlovic, PhD Head of Unit at PHENOMIN ‐ Institut Clinique de la Souris email: [email protected] CMC Strategy Forum Europe linkedIn: http://bit.ly/2VMLoJ9 Understanding Human disease with rodents: three challenges phenotypes of protein Some failure of mice and other coding genes in Mouse model organisms studies to be replicated or translated to humans
Multiple phenotypes No known Knowledge Translation phenotype on disease from rodent One genes to Human phenotype
Probability of success Reproducibility of preclinical studies https://www.mousephenotype.org/
50% of preclinical data are irreproducible CRISPR/Cas9 genome editing in rodents
THE GOOD What new possibilities does CRISPR open to better mimic human diseases ? – The genetic context (background) is not adapted – The designed mutation does not mimic the Human pathology
With CRISPR/Cas9 new types of mutations can be easily engineered
THE BAD There is more than off targets ! Understanding CRISPR/Cas9 genome editing to reduce lack of reproducibility – You are facing experimental variability, poor experimental design or bad reproducibility why do the results discovered using in vivo models sometimes fail to translate to human disease?
Plenty of literature showing that the inbred genetic background has an effect on the phenotype
Cerebellar phenotype of En1Hd/Hd mutants on 129/Sv and C57BL/6J backgrounds.
deletion of the cerebellum
129/Sv—En1Hd/Hd C57BL/6J—En1Hd/Hd Bilovocky et al. J. Neurosci. 2003 why do the results discovered using in vivo models sometimes fail to translate to human disease?
Plenty of literature showing that the inbred genetic background has an effect on the phenotype
genetic background limits generalizability of genotype-phenotype relationships Sittig et al. demonstrated low generalizability of mouse null allele phenotypes across a panel of F1 genetic backgrounds;
It suggests that the use of single strains is a barrier to robust characterization of genotype‐phenotype relationships.
Sittig et al. Neuron. 2016 Are you working on C57BL/6 or on C57BL/6 background ?
The extensive use of only few mouse strains like C57BL/6 cannot mimic the outbred diversity of human beings
In mice, C57BL/6 lines (congenic or coisogenic) represent 68% of the >28 000 lines available in MGI
Most of the phenotyping analyses were done in one of these genetic contexts, mixed backgrounds or 129* models were backcrossed to C57BL/6.
*129 mice are a complex collection of various backgrounds (Simpson et al., 1997)
C57BL/6 lines represent 68% of the >28 000 lines available (MGI extract, February 2017) Gene editing in animal disease model
NOD/ShiLtJ CRISPR/Cas9‐mediated gene editing for 10 genes delivered by microinjection
Gene X CRISPR/Cas9
NOD/ShiLtJ Gene X KO Qin et al. Genetics 2015
NOD/ShiLtJ: Non Obese diabetic, polygenic model for type 1 diabetes Manipulation of the Ts65Dn minichromosome
Ts65Dn mice carry a small extra chromosome derived from a reciprocal translocation (mouse chr16 and 17)
Partial trisomy model Ts65Dn Ts65Dn mice are trisomic for about two‐thirds of the genes orthologous to human chromosome 21. Chr17 Ts65Dn are well‐characterized and highly relevant model for studying Down Syndrome with neural cognitive deficits and behavioral abnormalities. Chr16 Manipulation of the Ts65Dn minichromosome
Partial trisomy model Ts65Dn
70 Mb are not related to Chr17 Down Syndrome disease
Chr16 Manipulation of the Ts65Dn minichromosome
Partial trisomy model Ts65Dn CRISPR /Cas9
Chr17 CRISPR /Cas9
Chr16 Manipulation of the Ts65Dn minichromosome
The Ts65Dn males show a decreased fertility We combined in vitro fertilization with CRISPR/Cas9 injection in eggs.
B6C3H xTs65Dn Understanding human structural variations leading to diseases
. Structural variants (SVs) are large genomic alterations that involve segments of DNA greater than one kb (Freeman et al., 2006) . Copy number variants (CNVs) are a subfamily of SV that correspond only to deletions or duplications and do not include inversions or translocations (Freeman et al., 2006)
. One of the most common cause in morbidity and mortality in human population . e.g. Down syndrome affecting 1 out of 800 births
. SVs likely play a major role in very various diseases not only restricted to neuronal disorders (Conrad et al., 2010; Fanciulli et al., 2007; McCarroll and Altshuler, 2007; Wu and Hurst, 2016)
. More than >60 000 SVs were discovered in human (Huddleston and Eichler, 2016) . The ones that are pathological SVs are mostly not known
. Chromosomalrearrangementshaveacentralroleinthepathogenesis of human cancers (Taki & Taniwaki, 2006) Generation of Cbs structural variant rat model
CRISPR/Cas9 injection in Sprague Dawley fertilized oocytes
4 gRNA used – 37.2 kb region
2 gRNAs 2 gRNAs
37.2 kb CRISPR/Cas9 for structural variant models ‐ Our results
CRISPR CRISPR
gene(s) Wild‐type
Deletion
Duplication (DUP) gene(s) gene(s)
Inversion (INV) gene(s)
DUP + INV gene(s) gene(s)
INV + DUP INV gene(s) gene(s)
INV + DUP gene(s) gene(s)
NHEJ gene(s) CRISPR/Cas9 make easy generating more complex mutations
Understanding the human structural variations that leads to disease
target DNA
deletion
– Working with a large panel of SV mutations both in mice and rat
– Read our paper Efficient and rapid generation of large genomic variants in rats and mice using CRISMERE ‐ Birling et al. 2017 Sci Rep. http://bit.ly/2Jlwxip Humanization of large locus in mice
Replacement of ApoE mouse gene by ApoE2, 3 and 4 human alleles including Tomm40 Humanization of large locus in mice
Replacement of ApoE mouse gene by ApoE2, 3 and 4 human alleles including Tomm40
Full replacement of a 37 kb sequence achieved in C57BL/6N ES cells using Neo/Hygro + CRISPR/Cas9 selection – 4 variants achieved
Human sequence 37.7 kb
Mouse locus
Mouse locus ~29 kb CRISPR/Cas9 genome editing can improve replicating Human disease
Generation of complex humanization in mouse and rat Achieving structural variants models Genetic context can be easily changed Working in a large panel of species
Read our review Modeling human disease in rodents by CRISPR/Cas9 genome editing. Birling et al. 2017 Mamm Genome. http://bit.ly/2WtsHYh Understating CRISPR/Cas9 genome editing to reduce reproducibility issues CRISPR/Cas9 is a (targeted) mutagen
DNA The only function of the CRISPR/Cas9 system is to create a double strand break at a chosen DNA sequence
CRISPR/Cas9 targeted double strand break (dsb) CRISPR/Cas9 is a (targeted) mutagen
DNA The only function of the CRISPR/Cas9 system is to create a double strand break at a chosen DNA sequence
CRISPR/Cas9 targeted double strand break (dsb)
The cell repairs the double strand break as it can NHEJ HR
Gene disruption Correction from by small the other insertions or chromosome deletions template The causes of unwanted mutations with CRISPR/Cas9
CRISPR/Cas9 is like bromide ethidium it is a mutagen
… a targeted mutagen but still a mutagen The causes of unwanted mutations with CRISPR/Cas9
CRISPR/Cas9 off‐target: – Lack of specificity of the system that cut DNA sequence that are similar to the target
Cell reparation off‐target: – Dsb are boosting cell reparation – Addition of CRISPR/Cas9 will favour unwanted (unanticipated) DNA reparation
Cell reparation on‐target: – Dsb are boosting cell reparation – Do I really have the model I think? CRISPR/Cas9 off‐target ‐ Lack of specificity of the system that cut DNA sequence that are similar to the target
Off‐target are very frequent Off‐target happens but are rare event
Cells In vivo ‐mice
Plasmid construct – constitutive expression mRNA or protein Cas9 – transient expression
+ IMPC data
(Iyer et al. 2015) CRISPR/Cas9 off‐target
With a good experimental design (CRISPR/Cas9 expressed as mRNA / protein for transient gene editing), off‐target level may be less important than natural genetic drift Generating a SNP model
DNA +
CRISPR/Cas9 targeted double strand break
NHEJ HR Cell reparation off‐targets
Single strand or double strand DNA used as template for recombination can integrate randomly in the genome Cell reparation off‐targets
Our data show that ssODN used to achieve knock‐in mutations can integrate at random in the genome >15% F0 animals have integrated random ssODN – > 100 animals screened from 5 projects Cell reparation off‐targets
Frequency of random integration of ssDNA need to be evaluated
Solutions – Random integration mutation can be removed by crossing with wt animals – qPCR screening for SNP (shortssDNA) or CKO / knock‐in (longssDNA) need to be performed Cell reparation on‐target
Reparation of the double strand break can lead to unexpected events
CRISPR/Cas9 mediate double strand‐ break may induce chromosomal rearrangements Lessons learned with CRISMERE
CRISPR CRISPR
Target
• The use of 2 or more sgRNAs may result in reparation of DNA leading to complex rearrangements like duplications, inversions or deletions. • By consequence those mutations should be anticipated. • Such rearrangements are frequent and have been observed for several loci both in mouse and rats. Translocations with CRISPR/Cas9
Two sgRNA in different genes = Two double strand breaks = 2 knock-out genes + Chromosomal Rearrangements Sequences rearrangements
Yin et al. proof of concepts of in vivo excision of HIV‐1 provirus in mouse
Yin et al., 2017
In Vivo Excision of HIV-1 Provirus by saCas9 and Multiplex Single-Guide RNAs in Animal Models Cell reparation on‐targets
CRISPR/Cas9 boost homologous recombination mechanisms HDR of a complex construct using CRISPR/cas
Classical ROSA26 HR arms: 15 kb circular targeting vector (6.9 kb Tg Knock-In)
Microinjection into C57BL/6N pronucleus
Circular plasmid Founder gRNA1 gRNA2 R5179‐ only one positive founder 52
gRNA1 gRNA2 WT
PHENOMIN‐ICS unpublished data Breeding scheme for founder 52
PCR Southern Blot 5 5 3 Internal 3’ Internal ’ ’ ’ F2 Extra +++ ++ Knock-in band PCR Southern Blot 5 5 3 Internal 3’ Internal PCR ’ ’ ’ F1 5 F2 Extra Internal 3’ +++ND ND ND +++ ++ ’ Knock-in Knock-in band
F2 Extra +++ ++ Knock-in band
F0 +++ Founder F1 Extra +++ ++ Knock-in band F2 Extra +++ ++ Knock-in band F1 Knock------out PHENOMIN‐ICS unpublished data Breeding scheme for founder 52
PCR Southern Blot 5 5 3 Internal 3’ Internal ’ ’ ’ F2 Extra +++ ++ Knock-in band PCR Southern Blot 5 5 3 Internal 3’ Internal PCR ’ ’ ’ F1 5 F2 Extra Internal 3’ +++ND ND ND +++ ++ ’ Knock-in Knock-in band
F2 Extra +++ ++ Knock-in band
F0 +++ Founder F1 Extra +++ ++ Knock-in band F2 Extra +++ ++ Knock-in band F1 Knock------out PHENOMIN‐ICS unpublished data Proposed mechanism
Integration of the transgene by homologous recombination (HR) followed by multiple integration of the vector by one crossing over
Target DNA circular HR vector Region of homology CRISPR Additional(s) recombination(s) event(s)
CO occur(s) using the HR vector as template
Transgene CRISPR/Cas9 mediated double strand break
Vector Transgene Transgene circular HR vector backbone
Transgene CO CO integration by HR Transgene
Region of homology CRISPR
PHENOMIN‐ICS unpublished data Validation by ddPCR
loxP loxP loxP loxP loxP loxP loxP loxP
ROSA26 ROSA26 transgene backbone transgene backbone transgene backbone transgene locus locus
PHENOMIN‐ICS unpublished data PHENOMIN‐ICS unpublished data Validation by ddPCR
loxP loxP loxP loxP loxP loxP loxP loxP
ROSA26 ROSA26 transgene backbone transgene backbone transgene backbone transgene locus locus
loxP
ROSA26 ROSA26 transgene locus locus
PHENOMIN‐ICS unpublished data Conclusion
We observed integration of circular DNA template as ‘concatemers’ at the target site
‘Concatemers’ are observed both in vitro (using CRISPR/Cas9 + circular donor in ES cells) or in vivo (CRISPR/Cas9 + circular donor pronuclear injection in eggs).
Frequency of this event is high in vitro when circular donor template is used : more than 50% of the mutant ES cells we tested showed concatemers integration.
We also observed ‘concatemers’ when using longssDNA as template ‘concatemers’ when using longssDNA
Long ssDNA 1,484 bp
5’ srPCR 3’ srPCR
291 bp 5' HR arm 929 bp Ex6 144 bp InterLoxP size 3' HR arm LoxP LoxP
External LR‐PCR
5’ srPCR External LR‐PCR
3‐ 2‐ 1.5‐ 1‐
3’ srPCR External LR‐PCR + sequencing Whole sequence correct ‘concatemers’ when using longssDNA
Long ssDNA 1,484 bp
ddPCR internal probe 5’ srPCR 3’ srPCR Avr II Avr II 291 bp 5' HR arm 929 bp Ex6 144 bp InterLoxP size 3' HR arm LoxP LoxP External LR‐PCR
External LR‐PCR Avr II 4.6 kb
Internal Southern probe
ddPCR ? 6‐ Unexpected 5‐ (2 copies ?) 4‐ WT (4.6 kb)
3‐ Copy Avr II digest Number 2 2.1 2.5 3.4 The causes of unwanted mutations with CRISPR/Cas9
CRISPR/Cas9 off‐target is maybe ‐ likely not the main concern when you are using CRISPR/Cas9 genome editing
Cell reparation mechanisms have a huge impact on off‐target & on‐target mutations Lost in translation
translation Clinical data How we use mouse model
• Hundreds people • Few small cohorts analysed • Outbred: genetic diversity • Inbred (C57BL/6): no genetic diversity • Strong impact of • Controlled environment environment • Non physiological • Physiological environment environment • Low reproducibility
few « reverse modelling » Member of the International Mouse Phenotyping Thank you for your attention Consortium to generate and phenotype each protein coding gene
PHENOMIN, ICS‐ Institut Clinique de la Souris Strasbourg, France
Expert in functional genomics (model Charles River & PHENOMIN alliance creation & phenotyping) for 17 years for model creation & phenotyping: collaborating with big pharma, European synergy of expertise & experience consortium and academic researchers