CRISPR/Cas9 in vivo dependency mapping reveals EZH2 as ... Engineering Thomas Naert.pdfDependency mapping in desmoid tumors via CRISPR/Cas9 Targeted amplicon sequencing of Dissect

CRISPR/Cas9 in vivo dependency mapping reveals EZH2 as druggable

target in desmoid tumors

Thomas Naert, PhDLab Kris Vleminckx – Ghent University

VIB Genome Engineering 2019

Left – znrf3/rnf43 TALENs, Right – rspo2 CRISPRSzenker-Ravi et al., Nature, 2018

Kindly shared by Dr. Amaya E.

rsp

o2

un

ilate

ral

Szenker-Ravi et al., 2018, nature

Restrictive CRISPR/Cas9 editing in X. tropicalis allows highly penetrant F0 disease models

Easily accessible and manipulatable

Diploid genome <-> laevis and zebrafish

Syntenic to the human genome

Tissue-restrictive targeted micro-injection

F0 mosaic mutant

rspo2CRISPR/Cas9

CRISPR-NSIDCRISPR/Cas9-mediated negative selection identification of dependencies

Expected INDEL editing outcomes

Machine learning prediction models

Experimental observations

Observed INDEL editing outcomes

In case of true dependenciesenrichment for in-frame mutations

Xenopus tropicalis OrganoidsZebrafish Genetic mice modelsCancer cell lines

Ascertaining negative selection in multiplex CRISPR/Cas9 experiments

Probability

theory

Pre-print available @

Ascertaining negative selection pressure in multiplex CRISPR/Cas9 experiments

The CRISPR/Cas9 INDEL scar

Time

What happens after a CRISPR/Cas9 cut?

Long standing idea that when a CRISPR/Cas9 cut is repaired without repair template,

That via NHEJ the CRISPR scarring pattern is:

66% frameshift mutations33% in-frame mutations

Due to the triplet coded nature of DNA

However, this paradigm is wrong…

Due to involvement of alternative NHEJ pathways,

ie. Micro-homology mediated end-joining

There is a gRNA-specific probability of frameshift and in-frame mutations,

Due to sequence context surrounding the gRNA cut site,

Every gRNA has a specific scarring pattern.

Outline

The outcomes of apc CRISPR/Cas9 editing and probabilistic modeling of CRISPR/Cas9 gene editing outcomes

Modeling of desmoid tumors in Xenopus tropicalis

CRISPR dependency mapping in desmoid tumors

Predictable template-free CRISPR/Cas9 repair in vertebrate embryos

Validating Ezh2 as a dependency factor in desmoid tumors

Outline






What happens after an apc CRISPR/Cas9 cut?

0

2

4

6

8

10

12

14

16

18

-28

-21

-20

-19

-18

-15

-14

-13

-11

-10 -8 -7 -5 -4 -3 -2 -1 1 2 3 4 5 7 8

11

13

Re

ads

wit

h s

pe

cifi

c a

pc

IND

EL v

aria

nt

(%)

Base Pair Changes

Miseq apc targeted amplicon sequencing

apc CRISPR/cas9

Normalized across 8 hearts – total of 3156 NGS readsMRV = mutant read variants

In-frame

Frameshift

Reduce

complexity

Apc CRISPR/Cas9 scarring pattern

0

10

20

30

40

50

60

70

80

90

in-frame MRV frameshift MRV

in-frame MRV

frameshift MRV

Apc CRISPR/Cas9 scarring pattern

50% chance on head

50% chance on tail

Normal Distribution

A quick pain-free refresh of your Probability Theory

CRISPR/Cas9 editing outcomes can be reduced to a ‘biased’ coin toss following binomial probabilistic theory

Two possible outcomes:

Frameshift (p)In-frame (1-p)

The CRISPR-Coin

“Heads” “Tails”

Introducing the CRISPR coin

Probabilistic modeling of CRISPR/Cas9 editing outcomes

0

20

40

60

80

100


in-frame MRV

frameshift MRV

apc CRISPR/Cas9 edit can be reduced to

Frameshift edit – 79% chance – “Heads”

In-frame edit – 21% chance – “Tails”

apc CRISPR/Cas9 scarring pattern

Success = frame-shift edit (Heads)n = # of apc allelles edited (Coins tossed)p = chance for obtaining a frameshift edit (chance on heads)

Outline






Desmoid tumor

Driven by Wnt signaling hyperactivation (apc or beta-catenin mutations)

Tumor of mesenchymal origin

Arise in deep muscle fascia, aponeurosis, and tendons

apc tumor suppressor mutations lead to desmoid tumors

apc CRISPR/Cas9

Biallelic mutant apc protein hyperactivates the Wnt pathwayand drives desmoid tumorigenesis

95% injected animals6 weeks old

apcmut/mut

apcmut/+

apc+/+

Positive selection

Editing of apc tumor suppressor leads to desmoid tumorigenesis in X. tropicalis

pbin7LEF:dGFP H&E

Frameshift (79%) In-frame (21%)

Frameshift (79%) =0,79 * 0,79 = 0,62 = 0,79 * 0,21 = 0,17

In-frame (21%) = 0,79 * 0,21 = 0,17 = 0,21 * 0,21 = 0,04

0

20

40

60

80

100


in-frame MRV

frameshift MRV

Sequence 83 desmoid tumorsAll are apc frameshift/frameshiftn = 83

Probabilistic modeling shows positive selection for frameshifting apc mutations in desmoid tumors

0

20

40

60

80

in-frame/in-frame frameshift/frameshift in-frame/frameshift

62%

heart tissue

Positive selectionNegative selection

79%21%

4% 34%

1 allele

2 alleles

apc CRISPR/Cas9

Success = biallelic apc frame-shift editn = # of tumorsp = chance on apc bialllelic frameshift edit

Heart apc scarring pattern

Expected tumor apc scarring pattern

Real-life tumor apc scarring pattern

Outline






SurgeryRadiotherapyClassical chemotherapy

High recurrence rate >50% within 2 years

While patient survival is high (>90% over 10 years)Chronic pain affects quality of life

Search for novel targets for targeted molecular treatmentDependency Factors

Aggressive treatments

Treatment modalities for desmoid tumors are inadequate

pre-surgery pre-surgery post-surgery

CRISPR library

Dependency factor

No dependency factor

Dependency factor screen – Dropout CRISPR/Cas9

CRISPR library

Dependency factor

No dependency factor

Negative selection

Dependency factor screen – Dropout CRISPR/Cas9

CRISPR-based in vivo identification of dependency factorsC

RIS

PR

/Cas

9 apc apc + not a dependency factor

apc + dependency factor

Legend

apc

Not dependency gene

Dependency gene

Wild-type

LOF mutations

In-frame functionalmutations

CR

ISP

R/C

as9 apc apc +

not a dependency factorapc +

dependency factor

Legend

apc

Not dependency gene

Dependency gene

Wild-type

LOF mutations


CRISPR-based in vivo identification of dependency factors

CR

ISP

R/C

as9 apc apc +

not a dependency factorapc +

dependency factorapc +

dependency factor

Legend

apc

Not essential gene

Essential gene

Wild-type

LOF mutations


For an dependency genebiallelic mutations are only possible ifminimum one allelle is in-frame and functional

CRISPR-based in vivo identification of dependency factors

Microarray8 desmoid tumors

22 ≠ fibrous lesions

RNA-seq7 desmoid tumors9 ≠ fibrous lesions

Genes identified in both datasets with FDR <1% and with contrast (fold change) > 2 in desmoid tumors

compared to other lesions

LOXADAM-12MDKHMMRWISP-1PYCR1

Prof. Matt van de RijnJoanna Przybyl

Stanford University

NUAK1FAP-αPCLAFEZH2CREB3L1

Potential desmoid tumor dependencies identified from clinical samples

Dependency mapping in desmoid tumors via CRISPR/Cas9

Targeted amplicon sequencing ofapc and potential dependency factorDissect tumor(s)

CRISPR/Cas9 apc + potential dependency factor

Biallelic frameshift mutations

Desmoid tumors can form with biallelic mutations in mdk and nuak1 and thus in abscence of Mdk and Nuak1 protein

0

2

4

6

8

10

12

14

16

18

20

ins1

de

l4

de

l8

de

l12

de

l16

de

l20

de

l24

de

l28

de

l32

de

l36

de

l40

de

l44

de

l48

de

l52

de

l56

ezh2_S692 inDelphi

ezh2_S692Experimental

0

2

4

6

8

10

12

14

16

18

20

ins1

de

l4

de

l8

de

l12

de

l16

de

l20

de

l24

de

l28

de

l32

de

l36

de

l40

de

l44

de

l48

de

l52

de

l56

ezh2_S692Experimental

0

2

4

6

8

10

12

14

16

18

20

ins1

de

l4

de

l8

de

l12

de

l16

de

l20

de

l24

de

l28

de

l32

de

l36

de

l40

de

l44

de

l48

de

l52

de

l56

ezh2_S692 inDelphi

Pearson correlation 0.819, p<0.001

Freq

uen

cy(%

)Freq

uen

cy (

%) Superimposition

Experimental scarring pattern of ezh2 gRNA

Freq

uen

cy(%

)

In silico predicted scarring pattern of ezh2 gRNA

ezh2 CRISPR/Cas9 scarring pattern to ascertain negative selection

Outline






Predictable template-free CRISPR/Cas9 editing outcomes in vertebrate embryos

C o rre la t io n s o f e x p e rim e n ta l d a ta to p re d ic tio n m o d e ls (n = 2 8 )

Pe

ars

on

co

rre

lati

on

Ind

elp

hi (m

ES

C)

Lin

del (K

562)

FO

RE

casT

(H

EK

293)

Ind

elp

hi (H

EK

293)

0 .0

0 .2

0 .4

0 .6

0 .8

1 .0

e zh 2 e d itin g in th e a d u lt h e a rt

v e rs u s In D e lp h i p re d ic t io n s

E x p e r im e n ta l o b s e r v a t io n o f v a r ia n t c a l l (% )

Pre

dic

ted

fr

eq

ue

nc

y o

f v

ar

ian

t c

all

(%

)

0 5 1 0 1 5 2 0

0

5

1 0

1 5

2 0r = 0 ,8 9 6 8

e z h 2 C R IS P R /C a s 9 g e n e e d it in g

o u tc o m e fin g e rp r in t

Fre

qu

en

cy

(%

)

inD

elp

hi

Pre

dic

ted

fre

qu

en

cy

of

var i

an

t call (

%)

Exp

er i

men

tal

Exp

er i

men

tal o

bserv

at i

on

of

var i

an

t call (

%)

0

5

1 0

1 5

2 0

In D e lp h i

E x p e rim e n ta l

P r e d ic t iv e p e r f o r m a n c e

E n d o g e n o u s ly o b s e rv e d

f r a m e s h i f t f r e q u e n c ie s

a m o n g a l l e d ite d p ro d u c ts in

X . t r o p ic a l is e m b r y o s ( % )

Pre

dic

te

d f

ra

me

sh

ift

fre

qu

en

cie

s

am

on

g a

ll m

ajo

r e

dit

ing

pro

du

cts

(%

)

0 2 0 4 0 6 0 8 0 1 0 0

0

2 0

4 0

6 0

8 0

1 0 0

In d e lp h i ( r = 0 .8 3 4 1 )

L in d e l ( r = 0 .7 0 9 3 )

F O R E c a s T ( r = 0 .6 4 9 1 )

F O R E c a s TP e a r s o n r = 0 ,6 3 1 4

E x p e r im e n ta l o b s e rv a t io n o f v a r ia n t c a ll (% )

Pre

dic

ted

fre

qu

en

cy

of

va

ria

nt

ca

ll (

%)

0 .1 1 1 0 1 0 0

0 .1

1

1 0

1 0 0

L in d e lP e a r s o n r = 0 ,5 4 4 5


Pre

dic

ted

fre

qu

en

cy

of

va

ria

nt

ca

ll (

%)

0 .1 1 1 0 1 0 0

0 .1

1

1 0

1 0 0

In d e lp h i (m E S C )P e a r s o n r = 0 ,8 5 5 8


Pre

dic

ted

fre

qu

en

cy

of

va

ria

nt

ca

ll (

%)

0 .1 1 1 0 1 0 0

0 .1

1

1 0

1 0 0

Naert et al; in preparation; 2020

Predictable template-free CRISPR/Cas9 editing outcomes in vertebrate embryos

Naert et al; in preparation; 2020

Indelphi-mESC model predicts CRISPR/Cas9 editing outcomes in:

Xenopus tropicalis

Xenopus laevisZebrafish

Outline







Expected INDEL editing outcomes

Machine learning prediction models

Experimental observations

Observed INDEL editing outcomes

In case of true dependenciesenrichment for in-frame mutations

Xenopus tropicalis OrganoidsZebrafish Genetic mice modelsCancer cell lines

Ascertaining negative selection in multiplex CRISPR/Cas9 experiments

Probability

theory

Pre-print available @

Ascertaining negative selection pressure in multiplex CRISPR/Cas9 experiments

e z h 2 C R IS P R /C a s 9 g e n e e d it in g

o u tc o m e fin g e rp r in t

Fre

qu

en

cy

(%

)

inD

elp

hi

Pre

dic

ted

fre

qu

en

cy

of

var i

an

t call (

%)

Exp

er i

men

tal

Exp

er i

men

tal o

bserv

at i

on

of

var i

an

t call (

%)

0

5

1 0

1 5

2 0

In D e lp h i

E x p e rim e n ta l

n = 4

Success = biallelic ezh2 frame-shift editn = # of tumorsp = chance on ezh2 bialllelic frameshift edit

Real-life tumor ezh2 scarring pattern

Negative selection for frameshifting ezh2 mutations in desmoid tumors

p (biallelic frameshift_experimental) = 0,57p (biallelic frameshift_Indelphi-mESC) = 0,52

r = 0.819 p<0.001

PRC2

H3K27 -----> H3K27Me3

EZH2 in-frame variants recovered in desmoid tumors remain functional

Targeting the catalytical site of ezh2 increases negative selection pressure

e z h 2 _ s 6 9 2 ta r g e ts th e e z h 2 c a ta ly t ic a l d o m a in

fre

qu

en

cy

of

fra

me

sh

ift

mu

tati

on

s (

%)

DT

1

DT

2

DT

3

DT

4

DT

5

DT

6

DT

7

DT

8

DT

9

DT

10

DT

11

DT

12

DT

13

DT

14

DT

15

DT

16

DT

17

DT

18

DT

19

DT

20

DT

21

DT

22

DT

23

DT

24

DT

25

DT

26

0 .0

0 .2

0 .4

0 .6

0 .8

1 .0

A P C

E Z H 2

Nature Biotechnology 2015

“Current screening strategies … often produces in-frame variants that retainfunctionality, which can obscure even strong genetic dependencies. Here weovercome this limitation by targeting CRISPR-Cas9 mutagenesis to exonsencoding functional protein domains. This generates a higher proportion ofnull mutations and substantially increases the potency of negative selection.”

CRISPR/Cas9apc + ezh2

The chance to observe this data, in absence of real negative selection, is smaller than 0,01%

Validation of the CRISPR-NSID – EZH2 as a dependency factor in desmoid tumors

Chemical inhibition via Tazemetostat

Selective Ezh2 inhibitor

Already advanced clinical

trials for other indications

Ezh2 inhibition as novel therapeutic strategy for established desmoid tumors

H3K27 ---> H3K27me3


Take Home

There is a gRNA-specific probability of frameshift and in-frame mutations – The scarring pattern

Due to sequence context surrounding the gRNA cut site

Predictable gene editing outcomes in vertebrate embryos via Indelphi-mESC model

Establishing these probabilities allows to determine deviations from the expected CRISPR/Cas9 scarring pattern in tumors

Positive selection for frameshift mutations in tumor supressors – apc

Negative selection for frameshift mutations in tumor dependencies – ezh2

In conclusion

Time

Wnt signaling ↑

apc mutations

Desmoid tumors in human patientA and B: pre-surgeryC: post-surgery recurrence

Ezh2 inhibition via Tazemetostat as potential new therapeutic approach for desmoid tumors

+ Tazemetostat

Acknowledgments

Unit Developmental BiologySuzan DemuynckDionysia DimitrakopoulouDieter TulkensMarjolein CarronCicekdal Munevver BurcuDr. Kris Vleminckx

Ex-membersDr. Tom Van NieuwenhuysenGriet Van ImschootDr. Rivka NoelandersDr. Hong Thi TranSven De GrandeRobin Colpaert

UZ Gent Pathology departmentDr. David Creytens

Stanford Clinical pathologyDr. Matt van de RijnDr. Joanna Przybyl

VIB protein core

And all other collaborators…

@XenoThomasNaert – Follow me on Twitter for all things CRISPR!

CRISPR/Cas9 in vivo dependency mapping reveals EZH2 as ... Engineering Thomas Naert.pdfDependency mapping in desmoid tumors via CRISPR/Cas9 Targeted amplicon sequencing of Dissect

Documents