RNA-based screening in drug discovery
Use of X-MAN™ isogenic cell lines in RNAi screening approaches
Comparison of siRNA and sgRNA screening approaches
The challenges of genome-wide CRISPR-Cas9 knockout (GeCKO) screening
Using CRISPR-Cas9 sgRNA for target identification and patient stratification
Moving from screening hit to target validation
sgRNA screening: not just KOs
RNA-based screening in drug discovery – introducing sgRNA technologies
1. 1
Translating Genomes | Personalizing Medicine
RNA-based screening in drug discovery – introducing sgRNA technologies
Dr. Jon Moore
CSO & VP Oncology
2. 2
Presenter
Jon Moore PhD
CSO & VP Oncology
Jon has been CSO of Horizon since February of this year after joining
Horizon in October 2012. His responsibilities include leading Horizon’s
portfolio of target ID & validation alliances and internal research portfolio.
Prior to this, he worked on multiple drug discovery programmes at Vernalis,
building on his extensive postdoctoral experience in the cell cycle field with
Tim Hunt and Sally Kornbluth.
His core areas of expertise are in oncology cell biology and alliance
management.
3. 3
Content of the Presentation
Introduction to Horizon Discovery
Horizon Discovery Research Division - Horizon’s services and collaboration platforms
RNA-based screening in drug discovery
• Use of X-MAN™ isogenic cell lines in RNAi screening approaches
• Comparison of siRNA and sgRNA screening approaches
• The challenges of genome-wide CRISPR-Cas9 knockout (GeCKO) screening
• Using CRISPR-Cas9 sgRNA for target identification and patient stratification
• Moving from screening hit to target validation
• sgRNA screening: not just KOs
Working with Horizon
5. 5
Horizon is a specialized drug discovery CRO, with a broad range of in vitro
and in vivo services
Drug Profiling, MOA
studies, Isogenic Cell Lines
Custom Isogenic Cell Line
Development
In vivo Models
Target Validation Studies,
Knockdown Assessment
High Throughput Drug
Combination
Screening Tumour
Microenvironment,
3D Cell Assays
Target Identification
siRNA/sgRNA screening
6. Use of X-MAN™ isogenic cell lines in RNAi
screening approaches
7. 7
SL screening in X-MANTM isogenics: Landscape of tumour mutations
Sequencing has identified 100’s of
potential driving mutations in cancer
Most “driver” mutations are rare and
poorly characterised
The most common driver mutations are in
tumour suppressors & can’t be drugged
directly
Certain frequently mutated oncogenes
(e.g. KRAS) are challenging to drug
Pan-cancer mutation rates (Tamberero et al., 2013)
Green boxes: druggable oncogenes
Red boxes: undruggable oncogenes & tumour suppressors
8. 8
SL screening in X-MANTM isogenics: Overcoming heterogeneity?
Where tumour suppressor loss is the cancer initiating
event, all tumour cells may share this mutation and any
vulnerabilities it confers
Agents exploiting this loss may assist with overcoming
tumour heterogeneity
Systematic de-orphaning is required to find key
druggable downstream targets: these could follow
either a co-dependence or a synthetic lethality
paradigm
9. 9
SL screening in X-MAN™ isogenics: On isogenic cell lines
Horizon has collection of over 550 isogenic cell line pairs X-MAN™ cell lines
Horizon’s expert targeting team can also engineer a genetically validated cell line to a
customer’s specification
10. 10
SL screening in X-MAN™ isogenics: Removing a key source of noise
X-MAN™ powered synthetic lethal screens
Exploits our genome editing technologies to make isogenic cells
Precisely control for target genotype & perfectly matched ‘normal’
Removes a key source of noise in SL-screens
Definitive; make any genotype of interest
siRNA and/or sgRNA libraries;
Or drug re-profiling
+
11. 11
SL screening in X-MANTM isogenics: Our siRNA screening library
Composed of 2,200 druggable genes (hand-selected) to high-interest oncology target classes.
Client/partner can request specific additions to the library
Dharmacon siGENOME smartpools used (4 siRNAs/target)
• Equimolar mixtures for enhanced on-target effects and reduced off-target events
12. 12
SL screening in X-MANTM isogenics: A typical waterfall plot
Three technical replicates per screen (CVs liquid handling <2.5%; CVs +/- controls ~ 13%)
Blue line is the median parental response plotted as a 2D waterfall.
Green data points are the mutant response for the matching target gene
Figure: Synthetic lethality waterfall plot (HCT116 isogenics)
Hits with selective death in
the mutant line
13. 13
SL screening in X-MAN™ isogenics: Hit confirmation workflow
siRNA Library Pooled Transfection
Lipid complexed with siRNA
siRNA > Lipid > Cells
Library stamped out 20n Transfection stamped out
6n pooled for mRNA
24, 48, 72 & 96hrs
16n pooled for protein
72hrs
WBqRT-PCR
KD and Biomarker analysis
AB staining for viability
Typically, we follow the following process:
1. Deconvolute SMARTpool into single siRNAs
2. Retest viability in isogenic pair; correlate
phenotype with degree of KD (RT-PCR &
immunoblots)
3. Assess breadth of synthetic lethality in cell line
panels (e.g. 5 lines with mutation vs. 5 lines w/o
mutation)
4. Rescue KD phenotype with siRNA-resistant cDNA
14. 14
SL screening in X-MAN™ isogenics: Complex assay systems
Our compact siRNA library is suitable for assessing target dependency under non-standard
conditions
Consider KRAS: in DLD1 cells KRASG13D is dispensable for 2D, but required for 3D growth
Activity of MEK, an effector of KRAS signalling is also required only under 3D
2D Adherent 3D Soft Agar
DLD-1 KRASG13D
DLD-1 KRASG13D
KO
Response to MEK inhibitors in 2D vs. 3D
Log [M] ARRY162 (MEK inhibitor)
15. 15
SL screening in X-MAN™ isogenics: Complex assay systems
Small siRNA library vs. literature KRAS SL targets screened in DLD1 cells under 2D vs 3D conditions
If 3D assay is more KRAS dependent, knockdown of these targets should have a greater effect
siRNA ranked by effect in 2D
Very anti-
proliferative
siRNAs
siRNA has
no effect on
growth
Positive controls
Negative
controls
Many siRNAs are more
anti-proliferative in 3D
KEY
17. 17
siRNA vs sgRNA: Limitations of siRNA/shRNA technology
Long experience with RNA interference highlights the following issues
1. Incomplete knockdown – leads to false negatives, especially when interrogating the function of
enzymes such as kinases
2. Off target effects e.g from SeeD sequence: especially problematic with shRNA
3. Duration of KD achievable with siRNA
shRNA vs KIF11 have different performances Various mechanisms for off-
target effects by shRNA
Transfection of siRNAs leads to partial KD of target expression
followed by time-dependent recovery in mRNA levels
Fellman & Lowe, Nat Cell Biol, 2014
http://www.broadinstitute.org/achilles
18. 18
siRNA vs sgRNA: Are KO screens better?
The shRNA infrastructure is easily adaptable to
knock-out screening
Lentiviruses can deliver Cas9 & sgRNAs into cells
with sufficient efficiency to perform whole
genome KO screens:
Custom sgRNA libraries can be readily assembled
in a pooled format after array based oligo
synthesis
Off-target effects don’t appear to be a major
concern
CAS 9
Guide RNA
PAM
sequence
Matching genomic
sequence
Genomic DNA
Shalem et al., Science 2013
19. 19
siRNA vs sgRNA: Are KO screens better?
Indications are Cas9/sgRNA suppresses gene
expression far more effectively than shRNA
+ve selection screens can ID genes whose
inactivation is required for survival
Where GFP expression is placed under control
of pathway reporters, +ve selection screens can
be run outside of oncology
Shalem et al., Science 2013
20. The challenges of genome-wide CRISPR-Cas9 knockout
(GeCKO) screening
21. 21
sgRNA screens: The challenges
-ve selection screens may be more problematic:
• sgRNAs efficacy is variable & KO efficiency likely depends on copy number:
• a typical sgRNA vs an essential gene is depleted ~ 4 fold from a 1n population, but only ~ 2-fold
from a 2n population
Second generation libraries may improve performance of –ve selection screens
Wang et al. (2014)
Build algorithm
22. 22
sgRNA screens: The challenges
Horizon intends to be at the forefront of guide design
• Multiple CRISPR/Cas9 experts are on SAB
• Relationship with Tech start-up Desktop Genetics who have built gUIDEbook™ tool for Horizon so that the
design algorithm can be updated
• We will assess the merits of guide design algorithms such as those presented by Doench et al. (2014)
Also exploring access to haploid cell lines to accelerate target ID
Prof. Feng Zhang
Broad Institute
CRISPR genome editing,
sgRNA screens
Prof. Keith Joung
Mass General Hospital
ZFN, TALEN, CRISPR
genome editing
Prof. Emmanuelle
Charpentier
Helmholtz Institute for
Infectious Disease
CRISPR/Cas9
24. 24
sgRNA screens: Evaluation of the technology at Horizon
Feng Zhang’s lab has improved the
lentiviral vector & sgRNA library
(see Sanjan et al., 2014)
v2 single vector system yields
approx. 7-fold higher titres
2-vector lentiviral system now
available for hard-to-infect cell lines
Horizon have the GeCKO v2 human
library in optimised single vector
system, plentiCRISPRv2:
• 6 sgRNAs/gene
• Targets miRNA
• 1000 non-targeting control
25. 25
sgRNA screens: Evaluation of the technology at Horizon
Horizon have also generated a 2344 sgRNA subset library containing 21 guides vs. each of
100+ genes including:
71 tumour suppressors described in Vogelstein et al. (2013)
29 literature targets that have been linked to synthetic lethality with KRAS mutations
Evaluation screens are being performed on ~ the following schedule
Our plans are to make a 2nd generation library targeting the 2200 genes in STX siRNA
library; currently selecting guide design algorithm
Sept 14 Oct14 Nov 14 Dec 14 Jan 15 Feb14Cell line Library Conditions
A375 TSG/KRAS SL Vemurafenib (R)
A375 TSG/KRAS SL 6-TG (R)
A375 TSG/KRAS SL Olaparib (S)
A375 TSG/KRAS SL R406 (S)
A375 GeCKO v2 Vemurafenib/Dabrafenib (R)
A375 GeCKO v2 Olaparib/R406
Undisclosed GeCKO Undisclosed (R)
Undisclosed GeCKO Undisclosed (S)
KRAS wt &
mutant
TSG/KRAS SL 2D & LA growth
Screens being
run together
Wet work at HDZ
NGS at GATC
Screens being
run together
Screens being
run together
26. 26
sgRNA screens: ID of vemurafenib resistance targets
Day -7
Transduce 64x106 cells with
TSG sgRNA library
(3 µl virus/2x106 cells)
Day -6 to 0
Select for transduced cells
with Puromycin
Day 1
Seed 24x106 cells/condition
in duplicate
(~10.000 cells/sgRNA)
Day 1 to 14
Treat cells with
PLX (2 µM) or DMSO
Harvest baseline timepoint
(Day 0) to check sgRNA
representation
Harvest end timepoint
(Day 14) to identify
sgRNA candidates
Screen readout
(~8.500x library coverage)
Cell numbers during PLX screen (D1 to D14)
31% transduced cells:
19.8x106 cells
(~8.500 cells/sgRNA)
TSG library contains sgRNAs vs
hit genes: NF1, NF2 for PLX
27. 27
sgRNA screens: ID of vemurafenib resistance targets
Average of normalized sgRNA counts from replicate cultures
>100-fold enrichment D14: DMSO vs PLX: 3 sgRNAs
(3x NF1: 1997, 1129, 1126)
>10-fold enrichment D14: DSMO vs PLX: 31 sgRNAs
(5x NF1, 5x NF2, 2x SMARCB1/MAPK9/ARID1A/SMAD4)
D0 D14_DMSO D14_PLX
28. 28
sgRNA screens: Enrichment of subset of NF1/NF2 sgRNAs by PLX treatment
>16-fold enrichment of NF1/NF2 sgRNAs
NF1 NF2
Log2ratio:sgRNAfrequencyPLXvsDMSO
Log2 ratios of highlighted sgRNAs
vs NF1/NF2 (in graph on the left):
>100x enrichment
29. 29
sgRNA screens: Log2 ratios of all NF1/NF2 sgRNAs after PLX treatment
NF1 NF2 NF1
4.8x
NF2
2.8x
Average
of all sgRNAs
Log2ratio:sgRNAfrequencyPLXvsDMSO
NF1 NF2
NF1 NF2
Average of sgRNAs from
different “backgrounds”:
Log2ratio:sgRNAfrequencyPLXvsDMSO
Log2ratio:sgRNAfrequencyPLXvsDMSO
31. 31
sgRNA screening follow up: Decision tree
How does one move an interesting hit, perhaps exemplified by just a single sgRNA, towards a
validated target?
The decision tree below outlines the kinds of steps Horizon would recommend
Is the hit exemplified by >1
sgRNA?
Repeat pooled screen with
more sgRNAs vs targets.
May want to include other
pathway elements
Move to well-based
validation
Is there doubt about the
penetrance of the
phenotype?
Assess whether target is
essential
Assess whether target
function is dependent on
its activity
Start drug discovery
Abandon target
NO
NO
NO
NO
YES
YES YES
SUCCESS
SUCCESS
YES
FAIL
FAIL
32. 32
sgRNA screening follow up: Pool & array based hit confirmation
sgRNA-rescreening
• Assembly of custom focussed pooled lentiviral
libraries with additional sgRNAs vs putative hits
is very cost-effective.
• Transduced libraries into multiple cell lines;
assess sgRNA-frequency by NGS after 14/21 days
Well-based validation
• sgRNA: first results in for arrayed expt.
A549 cells infected with one of 5 sgRNAs vs RAF1 or a
Rosa26 control
Puro selected for 7 days prior to plating
All 5 sgRNAs vs RAF1 reduce proliferation; expression of
Cas9 & sgRNA vs. Rosa26 still has phenotype
Further evaluation of controls & gene editing events
required
• Upgrading 2-vector v2 lentiCRIPSR system to
include fluorescent tags for rapid hit validation:
Shalem et al., Science 2013
33. 33
sgRNA screening follow up: Is my target essential?
We have devised a medium-throughput method that can shed light on the ambiguous results that
emerge from siRNA/shRNA or sgRNA screens
• shRNA only gives partial knockdown; growth phenotypes often partial too
• Repair of Cas9-mediated ds breaks can result in in-frame indels that don’t disrupt protein function
Use lentiviruses to deliver Cas9 + sgRNA vs target to cells, allow 14-20 days for gene editing to occur
and then culture colonies from single cells
Horizon’s cell-line engineering experience allows us to devise an analysis pipeline to characterise
editing events on a clonal level
Assess length of fluorescent PCR
products to check allele ratios and frame
shift occurrence in 100’s colonies
Culture colonies
from single cells
Infect cells
targeting gene
of interest
TARGET ESSENTIAL
Colonies contain only in
frame indels
TARGET NON-ESSENTIAL
Colonies contain frame shift
indels
34. 34
sgRNA follow up: Can an activity-dead allele of my target support its function?
• sgRNA + rAAV facilitates on-target engineering via HDR
• Very high editing efficiencies possible with selectable markers
• Multi-allele engineering likely in one step: Cas9/sgRNA will tend to cleave all alleles of targeted exon
• One or more alleles will be repaired by HDR; other allele(s) mis-repaired by NHEJ
• Horizon has a wealth of experience designing genome analysis protocols to ID desired clones
Horizon are exploring combining CRISPR/Cas9 nucleases with AAV donors to rapidly
generate multi-allele knock-in mutations
35. 35
sgRNA follow up: Can an activity-dead allele of my target support its function?
Data from a recent project where Horizon used CRISPR + AAV to introduce a minigene into a novel
target in diploid DLD1 cells
We have strong evidence for us having clones where one allele has been targeted by the AAV
cassette and the other mis-repaired by NHEJ
Guide# Cas-9 #wells with cells
# +ve for 5' and 3'
screen
single MT 224 29
single WT 415 177
dual MT 203 4
dual WT 336 43
Single guide + WT cas-9
5’ screening gel
Single guide + WT and MT cas-9
Amplification of non-targeted allele
* 518bp deletion
* *
* 524bp deletion
Summary of targeting events:
Cas9-wt + AAV minigene yields 42% on-target integrations
37. 37
Activity dead Cas9 mutants can be repurposed as transcriptional regulators
Experience with ZFNs & TALENs can quickly be applied to Cas9
• CRISPRi: Best inhibition of
expression comes from
sgRNAs vs nt 0-50 3’ to TSS
• CRISPRia: Best activation
of expression comes from
sgRNAs vs nt 150-50 5’ to
TSS
38. 38
Activity dead Cas9 mutants can be repurposed as transcriptional regulators
CRISPRi
Technology has potential to replace shRNA.
Likely to have fewer issue with off-target effects
Incomplete knockdown may be a limitation: not clear yet
how this compares with shRNA
CRISPRa
Likely to have uses in compound deconvolution studies +
studies of resistance
Limitation is that can only boost expression of intact genes
Not clear whether it can overcome epigenetic
downregulation
Horizon plans to:
Clone mutant Cas9-chimeras into plenti-CRISPR system
Select small library of sgRNA for pilot study
Perform resistance/sensitisation screens with vemurafenib
& olaparib – results potentially available in April/May
KRAB
39. Repositioning
Patient
stratification
LOH2LTarget ValidationTarget ID
siRNA screens
sgRNA screens
TIDVAL alliances
KO to test
essentiality
Activity dead KI
mutations
Generation of isogenic cell lines MOA assays to support med chem
Compound profiling in
isogenic cells
Combination assays
Compound profiling in large
cell line panels
Target validation & early stage drug discovery
collaborations
Finding a development path for
stranded clinical assets
In vivo models
Working with Horizon: Collaborations & services available
39
Discovery Research
Services
Custom Cell Line
Development
CombinatoRx
Custom Screening
Services
In vivo models
40. Your Horizon Contact:
Horizon Discovery Group plc, Building 7100, Cambridge Research Park, Waterbeach, Cambridge, CB25 9TL, United Kingdom
Tel: +44 (0) 1223 655 580 (Reception / Front desk) Fax: +44 (0) 1223 862 240 Email: info@horizondiscovery.com Web: www.horizondiscovery.com
Dr. Jon Moore
CSO & VP Oncology
j.moore@horizondiscovery.com
+44 (0)1223 655580
Editor's Notes
Welcome and thank you for joining me today to talk about cell-based assays at Horizon Discovery.
In the webinar today I would like to introduce you to Horizon Discovery and describe our functional assays division
Focussing on how we use cell based assays to aid our customers and collaborators to perform first class oncology drug discovery.
Information is no longer a bottleneck, emphasis is shifting to the ‘what does it all mean’
Genome editing is enabling the promise of the genomic era to be realized in the form of novel therapeutics and diagnostics
It involves the capability to efficiently introduce targeted alterations into any specific gene in living cells