Genome Editing & Gene Therapy
Eric Kelsic
CEO Dyno Therapeutics
Genome Editing
and Gene Therapy
Eric Kelsic
CEO Dyno Therapeutics
Cambridge, MA
Impact.Tech Seminar
San Francisco, June 11, 2019
1
Outline of our seminar
I: Genome editing new tools
II: AAV gene therapy new cures
III: Genetic engineering the actual
IV: What next? the possible
2
Self introduction
3
Systems Biology PhD
w/ Roy Kishony
@ Harvard
Physics BS
@ Caltech
Postdoc
w/ George Church
@ Wyss Institute
CEO Dyno
Therapeutics
Puzzles
Programming Startups
AAV
The journey toward modern medicine
4
Galen’s surgeries based on animal
dissection, dissection of human body
is taboo
Hippocratic Oath
prohibits euthanasia and abortion,
prohibits surgery for kidney stones
(lithotomy), prohibits divulging of
medical secrets
Renaissance: human
dissections and modern
anatomy 1800
1600
400 BCE
200 CE
Pasteur, Koch,
Cell & Germ Theory
1900
Blood
transfusions
DNA
structure
2000
Human
genome
Millions of patients have untreatable genetic diseases
5
•10,000+ known
monogenetic diseases
•1% of births worldwide
•Causing 20-30% of infant
deaths in US
Not treatable with current small
molecule or biologic drugs
WHO, Berry et al. 1987
I: Genome editing
discovering new tools
6
Genetics 101
DNA à RNA à Protein
7
(AGCT) (ACGU) (ACDEFGHIKLMNPQRSTVWY)
Proteins
are cellular
machines
8
David Goodsell
The human genome is
~3 billion nucleotides
~20,000 genes and many
more non-coding regulatory
elements
9
Annunziato , Scitable, nature.com
Diverse gene sequences lead to different phenotypes
10
Gene malfunction causes genetic disease
11
Toward base editing: a case study in genome editing
12
Komor et al, Nature 2016
Bioprospecting: treasures in the biosphere
• Nature has invented
~1e42 different protein
sequences
• 1e42 upper bound
• 1e30 bacteria in the
world
• 1e3 genes/cell
• Each gene changes each
year for 1e9 years
13
see Dryden, J. R. Soc. Interface, 2008 David Goodsell
Genome editing pre-CRISPR
14
1990
TALEs
Zinc Finger Nucleases
2000
2010
Kim, Nat Rev Gen 2014; David Goodsell
Meganucleases
CRISPR-Cas is a bacterial immune system
using an RNA-guided endonuclease
15
Rath, Biochimie 2015
Spacer sequences from
invading phage are stored in
the CRISPR module
RNA guided complex identifies and cleaves
protospacer sequences from invading
phages, interfering with phage infection
There are a wide variety of
biochemical mechanisms
for CRISPR interference
16
Rath, Biochimie 2015
Cas9 sgRNA enables simple programmed genome editing
17
Jinek, Science 2012; see also Gasiunas, PNAS 2012
18
Kim, Nat Rev Gen 2014
Adapting Cas9 for human genome editing – adding a
nuclear localization sequence (NLS)
19
Mali, Science 2013
Cong, Science 2013
Jinek, eLife 2013
Applications of Cas9 for basic
research
20
Wang, Science 2014; Shalem, Science 2014
Multiplexed programmable genome
perturbations perhaps the most
impactful initial application of Cas9:
• Knockout
• Activation
• Repression
• Epigenetics…
How will human therapeutics be developed?
21
Pro Con
Germ cells Mutations are inherited Unknown risks of off-target
mutations, ethics
Somatic cells Mutations are not
inherited
Challenge of delivery
ex vivo Enables screening for
positives and removal of
negatives
Complicated manufacturing
and transplantation
procedures
in vivo Easier to manufacture and
administer
Challenge of low efficiency
editing, dangers of off-
target editing
Question#1Question#2
✗
✓
Industry snapshot (June 2019)
22
Editas
23
ex vivo
in vivo
in vivo
Intellia
24
in vivo
ex vivo
CRISPR Therapeutics
25
in vivo
ex vivo
Main challenges facing human therapeutics?
26
HIGH EFFICIENCY EDITING LOW OFF TARGET EDITING
Why base editing?
27
HIGH EFFICIENCY EDITING LOW OFF TARGET EDITING
Cas9 base editors
28
Komor, Nature 2016
This is the simplest version,
additional elements enable
increased efficiency
High efficiency precision editing with base editing
29
Komor, Nature 2016
Expanding the capabilities of base editors: new PAMs
30
Hu, Nature 2018
Expanding the capabilities of base editors: new bases
31
Gaudelli, Nature 2017
Beam Therapeutics
32
Frontiers in base editing
• Precision of edits
• Reduced off-target edits
• In vivo delivery
• Reduced size (smaller Cas) for use in AAV
• Or viruses/capsids with larger capacity
33
Summary
• Excitement around genome editing driven by potential to understand
and treat genetic diseases with high unmet need
• Cas9 is an RNA-guided endonuclease enabling programmable genome
cutting, nicking, binding, deletion, insertion, substitution, …
• Base editing enables precision substitutions
• Broad consensus around the benefits of somatic cell genome editing
therapies, uncertainty around when germline editing will be acceptable
• Genome editing therapeutics can be ex vivo cell therapy or in vivo gene
therapy
34
II: AAV gene therapy
creating new cures
35
Why gene therapy?
36
MOST DRUGS ARE INHIBITORS MANY GENETIC DISEASE ARE
CAUSED BY LACK OF THE PROTEIN,
SO THERE’S NOTHING TO INHIBIT
GENETIC MALFUNCTIONS ARE
CHALLENGING TO TREAT USING
SMALL MOLECULES
Principles of in vivo gene therapy
37
Gene Therapy DNA Payload
=
Delivery
+
Replace or fix the
malfunctioning gene
Why use viral capsids?
38
LIPID NANOPARTICLES ARE
MATERIALS
CAPSIDS ARE MACHINES
• Flexible membranes
• Variable stickiness, reacts to
chemical properties
• Piggyback on endogenous
cellular machines
• Combinatorial chemical
synthesis
• Strong shell
• Senses and responds to
environmental signals
• Manipulates endogenous
cellular machines
• Genetic encoding enables
multiplexing
Gene therapy: early years
39
Friedmann & Roblin, first
proposal for gene therapy
1972
1980
1990
First attempts to treat
humans in clinical trials
For ADA-SCID, ex vivo via
lentivirus delivery
Development of viral delivery
vectors: Adenovirus, Lentivirus
Trials begin for in vivo gene
delivery using Adenovirus
2000
Tragedy motivates shift toward safer delivery
40
Adeno-Associated Virus is a small non-replicative
virus (dependovirus genus)
41
Kotterman, Nat. Rev. Gen. 2014
Adapting AAV for gene delivery
42
Kotterman, Nat. Rev. Gen. 2014
Advances in AAV therapeutic development
43
Dunbar, Science 2018
Immune suppression for gene therapy
• Prednisone
• Prevents inflammation
• Prevents immune system from
eliminating cells with viral
capsids
• Transient dosing schedule
concurrent with AAV
administration
44
Gene therapy for Leber’s Congenital Amaurosis (LCA)
• Payload: RPE65 (Retinoid
Isomerohydrolase, makes
retinol)
• Results in loss of vision then
blindness during childhood
• Prevalence: 1 in 200,000
• Subretinal AAV2 delivery
• Dose: 1.5e11vg/eye in 0.3mL
• Prednisone 3 days prior and
7 days after injection
• Spark Therapeutics
45
Broadening the potential of AAV
with new serotypes
46
Gao, J Virol. 2004DiMattia, J Virol. 2012
AAV2, 4, 8 & 9
New tricks:
self-complementary vectors
47
Wang, Nat. Rev. Drug Disc. 2019
Gene therapy for Spinal Muscular Atrophy (SMA type 1)
• Payload: SMN1 (Survival motor
neuron protein)
• Results in deterioration of motor
neurons, loss of mobility and then
death by 2 years
• Prevalence: 1 in 10,000
• Intravenous AAV9 delivery, self-
complementary vector
• Dose: 2e14vg/kg in 10-20mL/kg
• Prednisone as needed to control
immune response
• Avexis
48
Scaling up to meet manufacturing challenges
49
Vigene
Baculovirus for
large scale production
Other systems using
herpesvirus transduction,
suspension cells,
…
Clinical results
• The Gene Doctors, LCA treatment (39:17-43:14)
• https://www.pbs.org/video/the-gene-doctors-i1dd2h/
• Spinal Muscular Atrophy Treatment at Nationwide Children's -- Brett & Paige
• https://www.youtube.com/watch?v=3FyainVRdRc
• Gene Therapy for SMA Type 1: Evelyn's Story
• https://www.youtube.com/watch?v=yRrqbvUv6gQ
50
Adeno-Associated Virus (AAV): 50 years
51
Wang, Nat. Rev. Drug Disc. 2019
Clinical snapshot
52
Wang, Nat. Rev. Drug Disc. 2019
53
Selected AAV trials
Wang, Nat. Rev. Drug Disc. 2019
In vivo genome editing with AAV
54
(Gene knockouts)
Gene therapy frontiers: payloads
• Increased expression
• Control of dynamics
• Reduced size for large genes
• Dual and multi-payloads
• Combinatorial therapies
55
Gene therapy frontiers: capsids
• Not all cells and organs are reachable
• Limitations of cost and manufacturing capacity
56
Summary
• In vivo gene therapy delivers a genetic payload into your body’s cells
• Viral capsids are efficient machines for in vivo delivery
• AAV capsids enables in vivo disease treatment
• LCA, SMA, DMD, and many more…
• Gene therapy field has consolidated around AAV technology
• Early gene therapies are a roadmap for translation of in vivo genome
editing therapies
• With innovations on payloads and delivery, curing all genetic diseases is
within reach in our lifetime
57
III: Genetic engineering
expansion of the actual
58
Genetic engineering: how to develop new therapies for
treating disease as quickly as possible?
• Domesticating and improving our genetic natural resources
59
How can we make AAV capsids better for gene therapy?
60
Efficient delivery to target cells and tissues
Reduced off-target delivery
Robust evasion of pre-existing immunity
Increased packaging size
High-titer manufacturing capabilities
Natural
capsids
✓
✗
✗
✗
✓
Ideal
✓
✓
✓
✓
✓
AAV sequence space is vast and mostly unexplored
Explored:
1-4 mutations
between serotypes:
100-300 mutations
AAV2
AAV5
AAV9
AAV8
61
Genetic engineering: recombinant DNA technology
Tools for DNA manipulation
• 1972: plasmid cloning to propagate DNA in bacteria
• 1977: Sanger sequencing
• 1980’s: Phosphoramidite DNA synthesis
• 1983: PCR to selectively amplify DNA
• 1975: Asilomar – voluntary moratorium on risky applications of
recombinant DNA technology
62
Genetic engineering: synthetic biology
New tools à New name
• Parallel DNA sequencing (NGS)
• Long read sequencing (Pacbio,
Oxford Nanopore)
• Short read sequencing (Illumina)
• Today: read 600B bases for $10k
• DNA synthesis
• Primer oligos
• Custom gene synthesis
• Parallel oligo synthesis (oligo pools)
• 250,000 oligos at 230-300nt
• 75M bases for $5-20k
63
What challenges face a genetic engineer?
• Proteins are complex
• At present we can’t simulate or model that
complexity with low-level atomic models
• Protein interactions are poorly characterized
• Parameters for mid-level systems biology
models are not accurately measured and many
interactions are unknown
• Many higher levels of complexity
• From cells to tissues to organs to bodies
64
MAPK, cellsignal.com
Well then how does
Nature engineer?
65
Influenza, David Goodsell
Evolution is an algorithm
• With a goal of quickly finding improved sequences
• In nature: without any information whatsoever about how these
sequences work
66
Nature’s algorithm: Darwinian evolution
67
Random
mutation
Natural
selection
Natural rates of mutation
& recombination
Boltzmann
function of fitness
Directed evolution uses artificial selection
68
Lab control over
rates of mutation
& recombination
Often defaults to
greedy selection
Random
mutation
Artificial
selection
Engineering AAV capsids with directed evolution
69
Wang, Nat. Rev. Drug Disc. 2019
Random
mutation
Artificial selection
(mice, cells, primates)
Random
recombination
Why is AAV capsid engineering so challenging?
• As number of mutations
increases, probability of viability
with random search decreases
(exponentially)
• And … AAV needs to be
optimized across multiple
properties
Distance from WT
1 2 3 4 5
Probabilityofviability
Distance from WT
1 2 3 4 5
Probabilityofviability
A&B&C
AB
C
70
Random AAV libraries are typically “shallow”
Serotype A Serotype B
100-300 mutations
1-4 changes (~1e18 mutants) 1-4 changes
What’s needed is an efficient algorithmic
exploration of this space
Most useful areas of sequence space are unexplored
71
New technologies enable Super-Darwinian algorithms
72
Fitness
Number of library members tested
Super
Darwinian
evolution
Darwinian
evolution
Building smarter libraries with direct DNA synthesis
73
Print any DNA sequence
Currently: batch size is ~250,000
sequences up to 300nt
Limitation in length due to low
tolerance for deletions
74
Open-loop
physicallayer
Mutation
Selection by
Enrichment
Mutation
Selection by
Enrichment
Round 1 Round 2
Closed-loop
physicallayer
DNA
Synthesis
datalayer
Library Design
Machine
Learning
Sequencing
Enrichment #3
Enrichment #1
Enrichment #2
Round 1, 2, …
Selection
New technologies enable closed-loop workflows
Balancing exploration and exploitation
75
Random
mutation
Artificial
selection
Optimize for exploration
vs exploitation depending
on how many rounds of
selection remain
Sampler
Modeling fitness landscapes with machine learning
76
Random
mutation
Artificial
selection
Machine learning models
evaluate a large number
of sequences
Sampler Model
evaluation
Understanding AAV fitness landscapes for capsid
engineering
confidential
An outcome-focused approach to protein
engineering
Mechanistic approach
Outcome-focused approach
confidential 78
Mechanistic
understanding
Experiment HypothesisResults
Machine
Learning
Experiment SequenceOutcomes
Experiment SequenceOutcomes
Experiment SequenceOutcomes
Experiment SequenceOutcomes
79
Dyno Therapeutics
Dynoing
At Dyno Therapeutics we are building a machine-guided
AAV capsid engineering platform
80
Founded 2018, VC funded, based in Cambridge, MA
In contrast to traditional random approaches,
Dyno is systematically searching sequence space
Before Dyno
Random
High throughput
DNA Synthesis
Enrichment via
passaging
High throughput
DNA Sequencing
Single property Multiple properties
Library design
Selection strategy
Optimization
Machine
Learning
81
DNA
synthesis
DNA
sequencing
Biological
assays
Y
Y
Y
Y
Y
Y
Capsid
packaging
Immune
evasion
In vivo
delivery
Tissue
specificity
Machine
learning
Dyno’s high-throughput screening platform
82
Selection
Wide search Deep search
0 mutations
(WT)
1 mutation
0
# mutations
1 2 3 4 5
(WT)
Dyno Platform: Two algorithmic approaches
83
All experiments and data from Harvard (2015-2018)
with George Church and collaborators
84
George
Church
Pierce
Ogden
Sam
Sinai
• All possible codon changes
• Insertions
• Substitutions
• Deletions
• Includes barcode replicates and
internal controls
• Testing ~200,000 mutant
genomes in a single animal
• Emphasis on maximizing
information gained in the first
round of experiments
0 mutations
(WT)
1 mutation
Dyno platform: Wide search
85
AAV Capsid
GeneMutation Barcode
Wide search: Maximizing information gained
from a single experiment through multiplexing
YY
Y
Y
Y
Y
Capsid
packaging
Immune
evasion
In vivo
delivery
Tissue
specificity
NGS NGS NGS NGS 86
Case Study: Production
Wide Search
confidential
Outside Inside
confidential 88
Rep Cap
AAPMAAP
new gene
Landscape reveals new AAV biology
Dyno platform: Deep search
0
Distance from WT
1 2 3 4 5
(WT)
90
• Explore far away from natural
capsids by introducing many
mutations in combination
• Emphasis on improving function
based upon information gained
from previous rounds of
screening
AAV2
Can we explore deep into sequence space?
Results: still viable at
10-15 mutations
away from WT
=
50-64% similarity
within library region
Deep search pilot
• 80,000 mutants
• 28 amino acid region
• ~10,000 random
• ~15,000 internal replicas
• 1-42 mutations
Deep Search
91
Deep Search accelerates AAV engineering
Deep search
library
Random
library
Probabilityofviablepackaging
Distance from WT
92
Deep Search
Deep Search accelerates AAV engineering
Deep search
library
Random
library
Probabilityofviablepackaging
Distance from WT
+ deep
learning
+ future
rounds
93
Deep Search
Frontiers: Machine-guided capsid engineering
94
Efficient delivery to target cells and tissues
Reduced off-target delivery
Robust evasion of pre-existing immunity
Increased packaging size
High-titer manufacturing capabilities
Natural
capsids
✓
✗
✗
✗
✓
Ideal
✓
✓
✓
✓
✓
All of
these will
be solved
Up to 6kb
for AAV
Frontiers in genetic engineering for therapeutics
• Payload engineering
• Cell and tissue specific promoters
• Control over expression dynamics
• Machine-guided engineering
• Exploration strategies
• Machine learning architectures
• Transfer learning from other proteins
• Integration of data from diverse properties
• Technology development
• Advanced methods for DNA assembly and sequencing
95
Summary
• Evolution is an algorithm
• Darwinian evolution is random mutation + natural selection
• Directed evolution means artificial selection
• New technologies enable Super-Darwinian evolution
• Dyno Therapeutics is developing a machine-guided capsid engineering
platform
• Now hiring RA’s, Software engineers and experts in AAV gene therapy
• Major challenges in delivery with AAV capsids will be solved soon
• Machine-guided approaches will soon dominate genetic engineering
96
IV: What next?
exploration of the possible
97
Focusing on predictions rather than opinions
• Genome editing and gene therapy give us the ability to cure genetic
diseases and improve health
• Therapies that improve health may also enhance capabilities
• Based on today’s evidence, what is likely to come next?
98
Questions to consider
• Will genome editing and gene therapy become mainstream components
of a healthy lifestyle?
• What are the ethical concerns?
• When will these treatments be safe?
• When will benefits outweigh risks?
• When will benefits and risks be distributed justly and equitably?
99
Technological revolutions
• Once radical technologies are
now common
• Agriculture, Writing, Metals,
Medicine, Machines, Computers,
Internet
• Once radical behaviors are now
acceptable
• Ethical concerns on risks and
resources often have technical
solutions
• Technical revolutions flourish
under the current model of
economic liberalism
100
1. Germ-line editing with present day tools will not
become mainstream
• The option for embryo selection make germline editing unnecessary
• Germline editing has risks
• Our individualist society will most likely not support experiments that
violate genetic agency
101
2. Somatic therapies will predominate
• Somatic gene therapies are already successful and many more are under
development
• Gene therapy technology will develop faster than human generation time
• There will be no advantage to “locking in” a genetic change using germ-line
genome editing
102
3. Somatic therapies will redefine healthy
• Mono-genetic diseases will be cured through embryo selection, genome
editing and gene therapy
• Treatments will expand to polygenic and non-genetic diseases
• Mainstream therapies will lower costs and reduce risks
• Our individualistic society will tolerate genetic experimentation as a
form of self expression
• DIY somatic gene therapies will pioneer new genetic lifestyles
• The free market will develop mainstream somatic gene therapies
103
4. Genome synthesis will gradually replace ex-vivo
genome editing technologies
Surgery is to small molecule drugs
as
ex vivo cell therapy is to in vivo gene therapy
104
making a small number of changes complete rewriting
Editing Synthesis
With time, editing will introduce more changes, until
eventually we may skip editing in favor of synthesis
But applications of whole genome synthesis
technologies will depend on many factors
• Whole genome synthesis obsoletes the function of the germ line
• Disruptive potential of whole genome synthesis is far beyond that of
germ-line genome editing
• Whether genome synthesis will replace sexual recombination is
challenging to predict
• When will society embrace somatic therapies aimed at lifestyle improvement?
• When will whole genome synthesis capabilities be developed?
105
5. Genetics will become self-expression
• Humanity has always been defined more by behaviors than by genetics
• Genome editing, gene therapy and whole-genome synthesis
technologies will make genomics a personal choice
106
How to get involved?
107
Advice for scientists
• Move to Cambridge/Boston
• Why?
• Power laws: the best people
want to work with the best
people
• Boston area = Highest
density of talent, capital and
biotech innovation in the
world
108
Especially in gene therapy and genome editing:
Most big Pharma have R&D headquarters,
likewise for CRISPR companies
Advice for entrepreneurs
• Major challenges limiting the high potential of these technologies:
• Payment
• Distribution
• Manufacturing
• Delivery
• Redose
• Packaging size
• Payload
• Control over dynamics
• Sensing and actuation
• Polygenic diseases and beyond
109
Advice for investors
• Plenty of opportunities will be revealed by learning the science
• ASGCT and ESGCT conferences are a good place to start
• More smaller bets on disruptive technologies (vs incremental
technologies)
110
Final conclusions
• Millions of patients have genetic diseases that are untreatable today
• Genome sequencing, genome editing and gene therapy technologies
give us the tools to cure these diseases
• Somatic therapies will predominate over germline editing
• Progress in treating mono-genetic diseases will lead to advanced
treatments for polygenic and non-genetic diseases
• In the long term, individual experimentation, market forces, reduced
cost and reduced risk will make cell therapy and gene therapy part of a
mainstream healthy lifestyle
111
Appendix
references and resources
112
Recommended reading
• Genome editing with CRISPR
• https://www.addgene.org/crispr/guide/
• Gene Therapy
• Wang et al., 2019, https://www.ncbi.nlm.nih.gov/pubmed/30710128
• Machine-guided capsid engineering
• Kelsic & Church, 2019, https://insights.bio/cell-and-gene-therapy-
insights/journal/articles/challenges-and-opportunities-of-machine-guided-
capsid-engineering-for-gene-therapy/
• Upcoming publications from Church lab: Ogden at al., Sinai et al., etc.
• Dyno Therapeutics, www.dynotx.com
113

Genome Editing & Gene Therapy by Eric Kelsic

  • 1.
    Genome Editing &Gene Therapy Eric Kelsic CEO Dyno Therapeutics
  • 2.
    Genome Editing and GeneTherapy Eric Kelsic CEO Dyno Therapeutics Cambridge, MA Impact.Tech Seminar San Francisco, June 11, 2019 1
  • 3.
    Outline of ourseminar I: Genome editing new tools II: AAV gene therapy new cures III: Genetic engineering the actual IV: What next? the possible 2
  • 4.
    Self introduction 3 Systems BiologyPhD w/ Roy Kishony @ Harvard Physics BS @ Caltech Postdoc w/ George Church @ Wyss Institute CEO Dyno Therapeutics Puzzles Programming Startups AAV
  • 5.
    The journey towardmodern medicine 4 Galen’s surgeries based on animal dissection, dissection of human body is taboo Hippocratic Oath prohibits euthanasia and abortion, prohibits surgery for kidney stones (lithotomy), prohibits divulging of medical secrets Renaissance: human dissections and modern anatomy 1800 1600 400 BCE 200 CE Pasteur, Koch, Cell & Germ Theory 1900 Blood transfusions DNA structure 2000 Human genome
  • 6.
    Millions of patientshave untreatable genetic diseases 5 •10,000+ known monogenetic diseases •1% of births worldwide •Causing 20-30% of infant deaths in US Not treatable with current small molecule or biologic drugs WHO, Berry et al. 1987
  • 7.
  • 8.
    Genetics 101 DNA àRNA à Protein 7 (AGCT) (ACGU) (ACDEFGHIKLMNPQRSTVWY)
  • 9.
  • 10.
    The human genomeis ~3 billion nucleotides ~20,000 genes and many more non-coding regulatory elements 9 Annunziato , Scitable, nature.com
  • 11.
    Diverse gene sequenceslead to different phenotypes 10
  • 12.
    Gene malfunction causesgenetic disease 11
  • 13.
    Toward base editing:a case study in genome editing 12 Komor et al, Nature 2016
  • 14.
    Bioprospecting: treasures inthe biosphere • Nature has invented ~1e42 different protein sequences • 1e42 upper bound • 1e30 bacteria in the world • 1e3 genes/cell • Each gene changes each year for 1e9 years 13 see Dryden, J. R. Soc. Interface, 2008 David Goodsell
  • 15.
    Genome editing pre-CRISPR 14 1990 TALEs ZincFinger Nucleases 2000 2010 Kim, Nat Rev Gen 2014; David Goodsell Meganucleases
  • 16.
    CRISPR-Cas is abacterial immune system using an RNA-guided endonuclease 15 Rath, Biochimie 2015 Spacer sequences from invading phage are stored in the CRISPR module RNA guided complex identifies and cleaves protospacer sequences from invading phages, interfering with phage infection
  • 17.
    There are awide variety of biochemical mechanisms for CRISPR interference 16 Rath, Biochimie 2015
  • 18.
    Cas9 sgRNA enablessimple programmed genome editing 17 Jinek, Science 2012; see also Gasiunas, PNAS 2012
  • 19.
  • 20.
    Adapting Cas9 forhuman genome editing – adding a nuclear localization sequence (NLS) 19 Mali, Science 2013 Cong, Science 2013 Jinek, eLife 2013
  • 21.
    Applications of Cas9for basic research 20 Wang, Science 2014; Shalem, Science 2014 Multiplexed programmable genome perturbations perhaps the most impactful initial application of Cas9: • Knockout • Activation • Repression • Epigenetics…
  • 22.
    How will humantherapeutics be developed? 21 Pro Con Germ cells Mutations are inherited Unknown risks of off-target mutations, ethics Somatic cells Mutations are not inherited Challenge of delivery ex vivo Enables screening for positives and removal of negatives Complicated manufacturing and transplantation procedures in vivo Easier to manufacture and administer Challenge of low efficiency editing, dangers of off- target editing Question#1Question#2 ✗ ✓
  • 23.
  • 24.
  • 25.
  • 26.
  • 27.
    Main challenges facinghuman therapeutics? 26 HIGH EFFICIENCY EDITING LOW OFF TARGET EDITING
  • 28.
    Why base editing? 27 HIGHEFFICIENCY EDITING LOW OFF TARGET EDITING
  • 29.
    Cas9 base editors 28 Komor,Nature 2016 This is the simplest version, additional elements enable increased efficiency
  • 30.
    High efficiency precisionediting with base editing 29 Komor, Nature 2016
  • 31.
    Expanding the capabilitiesof base editors: new PAMs 30 Hu, Nature 2018
  • 32.
    Expanding the capabilitiesof base editors: new bases 31 Gaudelli, Nature 2017
  • 33.
  • 34.
    Frontiers in baseediting • Precision of edits • Reduced off-target edits • In vivo delivery • Reduced size (smaller Cas) for use in AAV • Or viruses/capsids with larger capacity 33
  • 35.
    Summary • Excitement aroundgenome editing driven by potential to understand and treat genetic diseases with high unmet need • Cas9 is an RNA-guided endonuclease enabling programmable genome cutting, nicking, binding, deletion, insertion, substitution, … • Base editing enables precision substitutions • Broad consensus around the benefits of somatic cell genome editing therapies, uncertainty around when germline editing will be acceptable • Genome editing therapeutics can be ex vivo cell therapy or in vivo gene therapy 34
  • 36.
    II: AAV genetherapy creating new cures 35
  • 37.
    Why gene therapy? 36 MOSTDRUGS ARE INHIBITORS MANY GENETIC DISEASE ARE CAUSED BY LACK OF THE PROTEIN, SO THERE’S NOTHING TO INHIBIT GENETIC MALFUNCTIONS ARE CHALLENGING TO TREAT USING SMALL MOLECULES
  • 38.
    Principles of invivo gene therapy 37 Gene Therapy DNA Payload = Delivery + Replace or fix the malfunctioning gene
  • 39.
    Why use viralcapsids? 38 LIPID NANOPARTICLES ARE MATERIALS CAPSIDS ARE MACHINES • Flexible membranes • Variable stickiness, reacts to chemical properties • Piggyback on endogenous cellular machines • Combinatorial chemical synthesis • Strong shell • Senses and responds to environmental signals • Manipulates endogenous cellular machines • Genetic encoding enables multiplexing
  • 40.
    Gene therapy: earlyyears 39 Friedmann & Roblin, first proposal for gene therapy 1972 1980 1990 First attempts to treat humans in clinical trials For ADA-SCID, ex vivo via lentivirus delivery Development of viral delivery vectors: Adenovirus, Lentivirus Trials begin for in vivo gene delivery using Adenovirus 2000
  • 41.
    Tragedy motivates shifttoward safer delivery 40
  • 42.
    Adeno-Associated Virus isa small non-replicative virus (dependovirus genus) 41 Kotterman, Nat. Rev. Gen. 2014
  • 43.
    Adapting AAV forgene delivery 42 Kotterman, Nat. Rev. Gen. 2014
  • 44.
    Advances in AAVtherapeutic development 43 Dunbar, Science 2018
  • 45.
    Immune suppression forgene therapy • Prednisone • Prevents inflammation • Prevents immune system from eliminating cells with viral capsids • Transient dosing schedule concurrent with AAV administration 44
  • 46.
    Gene therapy forLeber’s Congenital Amaurosis (LCA) • Payload: RPE65 (Retinoid Isomerohydrolase, makes retinol) • Results in loss of vision then blindness during childhood • Prevalence: 1 in 200,000 • Subretinal AAV2 delivery • Dose: 1.5e11vg/eye in 0.3mL • Prednisone 3 days prior and 7 days after injection • Spark Therapeutics 45
  • 47.
    Broadening the potentialof AAV with new serotypes 46 Gao, J Virol. 2004DiMattia, J Virol. 2012 AAV2, 4, 8 & 9
  • 48.
  • 49.
    Gene therapy forSpinal Muscular Atrophy (SMA type 1) • Payload: SMN1 (Survival motor neuron protein) • Results in deterioration of motor neurons, loss of mobility and then death by 2 years • Prevalence: 1 in 10,000 • Intravenous AAV9 delivery, self- complementary vector • Dose: 2e14vg/kg in 10-20mL/kg • Prednisone as needed to control immune response • Avexis 48
  • 50.
    Scaling up tomeet manufacturing challenges 49 Vigene Baculovirus for large scale production Other systems using herpesvirus transduction, suspension cells, …
  • 51.
    Clinical results • TheGene Doctors, LCA treatment (39:17-43:14) • https://www.pbs.org/video/the-gene-doctors-i1dd2h/ • Spinal Muscular Atrophy Treatment at Nationwide Children's -- Brett & Paige • https://www.youtube.com/watch?v=3FyainVRdRc • Gene Therapy for SMA Type 1: Evelyn's Story • https://www.youtube.com/watch?v=yRrqbvUv6gQ 50
  • 52.
    Adeno-Associated Virus (AAV):50 years 51 Wang, Nat. Rev. Drug Disc. 2019
  • 53.
    Clinical snapshot 52 Wang, Nat.Rev. Drug Disc. 2019
  • 54.
    53 Selected AAV trials Wang,Nat. Rev. Drug Disc. 2019
  • 55.
    In vivo genomeediting with AAV 54 (Gene knockouts)
  • 56.
    Gene therapy frontiers:payloads • Increased expression • Control of dynamics • Reduced size for large genes • Dual and multi-payloads • Combinatorial therapies 55
  • 57.
    Gene therapy frontiers:capsids • Not all cells and organs are reachable • Limitations of cost and manufacturing capacity 56
  • 58.
    Summary • In vivogene therapy delivers a genetic payload into your body’s cells • Viral capsids are efficient machines for in vivo delivery • AAV capsids enables in vivo disease treatment • LCA, SMA, DMD, and many more… • Gene therapy field has consolidated around AAV technology • Early gene therapies are a roadmap for translation of in vivo genome editing therapies • With innovations on payloads and delivery, curing all genetic diseases is within reach in our lifetime 57
  • 59.
  • 60.
    Genetic engineering: howto develop new therapies for treating disease as quickly as possible? • Domesticating and improving our genetic natural resources 59
  • 61.
    How can wemake AAV capsids better for gene therapy? 60 Efficient delivery to target cells and tissues Reduced off-target delivery Robust evasion of pre-existing immunity Increased packaging size High-titer manufacturing capabilities Natural capsids ✓ ✗ ✗ ✗ ✓ Ideal ✓ ✓ ✓ ✓ ✓
  • 62.
    AAV sequence spaceis vast and mostly unexplored Explored: 1-4 mutations between serotypes: 100-300 mutations AAV2 AAV5 AAV9 AAV8 61
  • 63.
    Genetic engineering: recombinantDNA technology Tools for DNA manipulation • 1972: plasmid cloning to propagate DNA in bacteria • 1977: Sanger sequencing • 1980’s: Phosphoramidite DNA synthesis • 1983: PCR to selectively amplify DNA • 1975: Asilomar – voluntary moratorium on risky applications of recombinant DNA technology 62
  • 64.
    Genetic engineering: syntheticbiology New tools à New name • Parallel DNA sequencing (NGS) • Long read sequencing (Pacbio, Oxford Nanopore) • Short read sequencing (Illumina) • Today: read 600B bases for $10k • DNA synthesis • Primer oligos • Custom gene synthesis • Parallel oligo synthesis (oligo pools) • 250,000 oligos at 230-300nt • 75M bases for $5-20k 63
  • 65.
    What challenges facea genetic engineer? • Proteins are complex • At present we can’t simulate or model that complexity with low-level atomic models • Protein interactions are poorly characterized • Parameters for mid-level systems biology models are not accurately measured and many interactions are unknown • Many higher levels of complexity • From cells to tissues to organs to bodies 64 MAPK, cellsignal.com
  • 66.
    Well then howdoes Nature engineer? 65 Influenza, David Goodsell
  • 67.
    Evolution is analgorithm • With a goal of quickly finding improved sequences • In nature: without any information whatsoever about how these sequences work 66
  • 68.
    Nature’s algorithm: Darwinianevolution 67 Random mutation Natural selection Natural rates of mutation & recombination Boltzmann function of fitness
  • 69.
    Directed evolution usesartificial selection 68 Lab control over rates of mutation & recombination Often defaults to greedy selection Random mutation Artificial selection
  • 70.
    Engineering AAV capsidswith directed evolution 69 Wang, Nat. Rev. Drug Disc. 2019 Random mutation Artificial selection (mice, cells, primates) Random recombination
  • 71.
    Why is AAVcapsid engineering so challenging? • As number of mutations increases, probability of viability with random search decreases (exponentially) • And … AAV needs to be optimized across multiple properties Distance from WT 1 2 3 4 5 Probabilityofviability Distance from WT 1 2 3 4 5 Probabilityofviability A&B&C AB C 70
  • 72.
    Random AAV librariesare typically “shallow” Serotype A Serotype B 100-300 mutations 1-4 changes (~1e18 mutants) 1-4 changes What’s needed is an efficient algorithmic exploration of this space Most useful areas of sequence space are unexplored 71
  • 73.
    New technologies enableSuper-Darwinian algorithms 72 Fitness Number of library members tested Super Darwinian evolution Darwinian evolution
  • 74.
    Building smarter librarieswith direct DNA synthesis 73 Print any DNA sequence Currently: batch size is ~250,000 sequences up to 300nt Limitation in length due to low tolerance for deletions
  • 75.
    74 Open-loop physicallayer Mutation Selection by Enrichment Mutation Selection by Enrichment Round1 Round 2 Closed-loop physicallayer DNA Synthesis datalayer Library Design Machine Learning Sequencing Enrichment #3 Enrichment #1 Enrichment #2 Round 1, 2, … Selection New technologies enable closed-loop workflows
  • 76.
    Balancing exploration andexploitation 75 Random mutation Artificial selection Optimize for exploration vs exploitation depending on how many rounds of selection remain Sampler
  • 77.
    Modeling fitness landscapeswith machine learning 76 Random mutation Artificial selection Machine learning models evaluate a large number of sequences Sampler Model evaluation
  • 78.
    Understanding AAV fitnesslandscapes for capsid engineering confidential
  • 79.
    An outcome-focused approachto protein engineering Mechanistic approach Outcome-focused approach confidential 78 Mechanistic understanding Experiment HypothesisResults Machine Learning Experiment SequenceOutcomes Experiment SequenceOutcomes Experiment SequenceOutcomes Experiment SequenceOutcomes
  • 80.
  • 81.
    At Dyno Therapeuticswe are building a machine-guided AAV capsid engineering platform 80 Founded 2018, VC funded, based in Cambridge, MA
  • 82.
    In contrast totraditional random approaches, Dyno is systematically searching sequence space Before Dyno Random High throughput DNA Synthesis Enrichment via passaging High throughput DNA Sequencing Single property Multiple properties Library design Selection strategy Optimization Machine Learning 81
  • 83.
  • 84.
    Wide search Deepsearch 0 mutations (WT) 1 mutation 0 # mutations 1 2 3 4 5 (WT) Dyno Platform: Two algorithmic approaches 83
  • 85.
    All experiments anddata from Harvard (2015-2018) with George Church and collaborators 84 George Church Pierce Ogden Sam Sinai
  • 86.
    • All possiblecodon changes • Insertions • Substitutions • Deletions • Includes barcode replicates and internal controls • Testing ~200,000 mutant genomes in a single animal • Emphasis on maximizing information gained in the first round of experiments 0 mutations (WT) 1 mutation Dyno platform: Wide search 85
  • 87.
    AAV Capsid GeneMutation Barcode Widesearch: Maximizing information gained from a single experiment through multiplexing YY Y Y Y Y Capsid packaging Immune evasion In vivo delivery Tissue specificity NGS NGS NGS NGS 86
  • 88.
    Case Study: Production WideSearch confidential
  • 89.
  • 90.
    Rep Cap AAPMAAP new gene Landscapereveals new AAV biology
  • 91.
    Dyno platform: Deepsearch 0 Distance from WT 1 2 3 4 5 (WT) 90 • Explore far away from natural capsids by introducing many mutations in combination • Emphasis on improving function based upon information gained from previous rounds of screening
  • 92.
    AAV2 Can we exploredeep into sequence space? Results: still viable at 10-15 mutations away from WT = 50-64% similarity within library region Deep search pilot • 80,000 mutants • 28 amino acid region • ~10,000 random • ~15,000 internal replicas • 1-42 mutations Deep Search 91
  • 93.
    Deep Search acceleratesAAV engineering Deep search library Random library Probabilityofviablepackaging Distance from WT 92 Deep Search
  • 94.
    Deep Search acceleratesAAV engineering Deep search library Random library Probabilityofviablepackaging Distance from WT + deep learning + future rounds 93 Deep Search
  • 95.
    Frontiers: Machine-guided capsidengineering 94 Efficient delivery to target cells and tissues Reduced off-target delivery Robust evasion of pre-existing immunity Increased packaging size High-titer manufacturing capabilities Natural capsids ✓ ✗ ✗ ✗ ✓ Ideal ✓ ✓ ✓ ✓ ✓ All of these will be solved Up to 6kb for AAV
  • 96.
    Frontiers in geneticengineering for therapeutics • Payload engineering • Cell and tissue specific promoters • Control over expression dynamics • Machine-guided engineering • Exploration strategies • Machine learning architectures • Transfer learning from other proteins • Integration of data from diverse properties • Technology development • Advanced methods for DNA assembly and sequencing 95
  • 97.
    Summary • Evolution isan algorithm • Darwinian evolution is random mutation + natural selection • Directed evolution means artificial selection • New technologies enable Super-Darwinian evolution • Dyno Therapeutics is developing a machine-guided capsid engineering platform • Now hiring RA’s, Software engineers and experts in AAV gene therapy • Major challenges in delivery with AAV capsids will be solved soon • Machine-guided approaches will soon dominate genetic engineering 96
  • 98.
    IV: What next? explorationof the possible 97
  • 99.
    Focusing on predictionsrather than opinions • Genome editing and gene therapy give us the ability to cure genetic diseases and improve health • Therapies that improve health may also enhance capabilities • Based on today’s evidence, what is likely to come next? 98
  • 100.
    Questions to consider •Will genome editing and gene therapy become mainstream components of a healthy lifestyle? • What are the ethical concerns? • When will these treatments be safe? • When will benefits outweigh risks? • When will benefits and risks be distributed justly and equitably? 99
  • 101.
    Technological revolutions • Onceradical technologies are now common • Agriculture, Writing, Metals, Medicine, Machines, Computers, Internet • Once radical behaviors are now acceptable • Ethical concerns on risks and resources often have technical solutions • Technical revolutions flourish under the current model of economic liberalism 100
  • 102.
    1. Germ-line editingwith present day tools will not become mainstream • The option for embryo selection make germline editing unnecessary • Germline editing has risks • Our individualist society will most likely not support experiments that violate genetic agency 101
  • 103.
    2. Somatic therapieswill predominate • Somatic gene therapies are already successful and many more are under development • Gene therapy technology will develop faster than human generation time • There will be no advantage to “locking in” a genetic change using germ-line genome editing 102
  • 104.
    3. Somatic therapieswill redefine healthy • Mono-genetic diseases will be cured through embryo selection, genome editing and gene therapy • Treatments will expand to polygenic and non-genetic diseases • Mainstream therapies will lower costs and reduce risks • Our individualistic society will tolerate genetic experimentation as a form of self expression • DIY somatic gene therapies will pioneer new genetic lifestyles • The free market will develop mainstream somatic gene therapies 103
  • 105.
    4. Genome synthesiswill gradually replace ex-vivo genome editing technologies Surgery is to small molecule drugs as ex vivo cell therapy is to in vivo gene therapy 104 making a small number of changes complete rewriting Editing Synthesis With time, editing will introduce more changes, until eventually we may skip editing in favor of synthesis
  • 106.
    But applications ofwhole genome synthesis technologies will depend on many factors • Whole genome synthesis obsoletes the function of the germ line • Disruptive potential of whole genome synthesis is far beyond that of germ-line genome editing • Whether genome synthesis will replace sexual recombination is challenging to predict • When will society embrace somatic therapies aimed at lifestyle improvement? • When will whole genome synthesis capabilities be developed? 105
  • 107.
    5. Genetics willbecome self-expression • Humanity has always been defined more by behaviors than by genetics • Genome editing, gene therapy and whole-genome synthesis technologies will make genomics a personal choice 106
  • 108.
    How to getinvolved? 107
  • 109.
    Advice for scientists •Move to Cambridge/Boston • Why? • Power laws: the best people want to work with the best people • Boston area = Highest density of talent, capital and biotech innovation in the world 108 Especially in gene therapy and genome editing: Most big Pharma have R&D headquarters, likewise for CRISPR companies
  • 110.
    Advice for entrepreneurs •Major challenges limiting the high potential of these technologies: • Payment • Distribution • Manufacturing • Delivery • Redose • Packaging size • Payload • Control over dynamics • Sensing and actuation • Polygenic diseases and beyond 109
  • 111.
    Advice for investors •Plenty of opportunities will be revealed by learning the science • ASGCT and ESGCT conferences are a good place to start • More smaller bets on disruptive technologies (vs incremental technologies) 110
  • 112.
    Final conclusions • Millionsof patients have genetic diseases that are untreatable today • Genome sequencing, genome editing and gene therapy technologies give us the tools to cure these diseases • Somatic therapies will predominate over germline editing • Progress in treating mono-genetic diseases will lead to advanced treatments for polygenic and non-genetic diseases • In the long term, individual experimentation, market forces, reduced cost and reduced risk will make cell therapy and gene therapy part of a mainstream healthy lifestyle 111
  • 113.
  • 114.
    Recommended reading • Genomeediting with CRISPR • https://www.addgene.org/crispr/guide/ • Gene Therapy • Wang et al., 2019, https://www.ncbi.nlm.nih.gov/pubmed/30710128 • Machine-guided capsid engineering • Kelsic & Church, 2019, https://insights.bio/cell-and-gene-therapy- insights/journal/articles/challenges-and-opportunities-of-machine-guided- capsid-engineering-for-gene-therapy/ • Upcoming publications from Church lab: Ogden at al., Sinai et al., etc. • Dyno Therapeutics, www.dynotx.com 113