Reasonably Random Synthetic Biology at Amyris

Tim GardnerDirector, Research Programs & Operations

October 27, 2010

2

Overview

► Amyris is an integrated renewable products company producing advanced renewable fuels and chemicals

► Founded in 2003 on principle of social responsibility: use our know-how to address biggest health and environmental challenges

► Public company (IPO September 2010) with R&D, Manufacturing and Distribution facilities in the Emeryville, CA, Campinas, Brazil & Chicago, IL

3

Artemisinin is 95% effective against malariaThe Challenge: Supplying Artemisinin Anti-Malarials

50X increase in production10X decrease in price

Treating malaria would require:300 to 500 million treatments per year

Artemisinin treatments needed:225 to 400 tons of artemisinin per year

This would require:6,000,000 tons of plant material

Malaria causes:1 to 3 million deaths per year

Total Chemical Synthesis too expensive

Amyris’ fouding product: Artemsinin

4

ispA

G6P

FDP

G3P

PEP

PYR

AcCoA

OAA

MAL

CIT

IPP

TCACycle

FPP

idiDMAP

Glucose

MevalonatePathway

Amorphadiene(arteminin precurser)

Artemisaannua

steroidsquinonesmembranes

Non-profit effort to manufacture Artemsinin

5

Strain performance targets reached

0

10

20

30

40

50

0.00 20.00 40.00 60.00 80.00 100.00 120.00 140.00

Am

orph

adie

ne [g

/L]

Time (hrs)

Improvement 1Improvement 2Improvement 3Improvement 4

25g/L target

• Sanofi-aventis now ramping production, formulation and product stability testing• Aim for world-wide distribution in 2012.

6

From drugs to fuels

Phase-contrast micrograph of Amyris engineered microbes producing precursor to Amyris Renewable Diesel

Isoprenoid technology platform capable of making more than 50,000 molecules

• Hydrocarbons, not alcohols or esters

• Can be used in existing engines with no performance trade-offs

• Superior environmental profile– substantially lower greenhouse

gas emissions than petroleum– No sulfur– Lower particulates and NOx

• Can be delivered using existing distribution infrastructure

7

ispA

G6P

FDP

G3P

PEP

PYR

AcCoA

OAA

MAL

CIT

IPP

TCACycle

FPP

idiDMAP

Glucose

Farenesene

steroidsquinonesmembranes

MevalonatePathway

Diesel production

8Note: Amyris diesel will be used in blends with conventional fuels; values shown for Amyris diesel is for our biomass derived blending component; SME = Soy Methyl Esters

+1

< – 50 AmyrisDiesel

FAME

# 2-D

47

58.1

40-55

Cetane Number

AmyrisDiesel

FAME

# 2-D

Energy Density1000 BTU/gal

118

121

115-142

AmyrisDiesel

FAME

# 2-D

0 50 100 1500 20 40 60-75 -50 -25 0

-9 to-30

Cloud Point (°C)(cold temp operation)

Additional benefits of Amyris renewable diesel compared to #2-Diesel• 90%+ lower greenhouse gas emissions• No sulfur & produces lower NOx and particulate emissions• Registered with the EPA for 20% blends

Amyris Renewable Diesel: a better fuel

9

>$1 Trillion dollar market accessible

fermentation

• Consumer products– detergents– Cosmetics– fragrances

• Lubricants– family of base oils – designed to be high

performance

• Polymers– adhesives– oxygen scavenger– toughening agent

• Renewable diesel– “plug-in” fuels – meets or exceeds stds– substantially lower

emissions

• Other applications– crop protection– many others

Farnesene

By combining biology and chemistry, Biofenebecomes a building block of renewable products for a diverse set of applications

biology

chemistry

$48B

$337B

$809B

>$50B

10

Lower cost of production enables access to larger markets

11

Low-cost production drives everything in strain R&D

Short term Medium term Long term

Time to value

Fam

iliar

Unf

amili

arU

ncer

tain

Leve

l of r

isk

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

Capital cost-saving opportunities

Engineering decisions drive strain performance criteria

Multi-parameter strain optimization problem• yield• productivity• reduced media supplements• temperature• biocatalyst stability• GMM certification

Amount of savings

For fuel synthesis we aim to direct >90% of cell resources to the synthesis of byproducts under stringent productivity, temperature, and media conditions

2M ton/yr plant

12

like getting a toddler to eat salad

13

Apr

May

JunJulAug

SepOct

Nov

Dec

JanFebMar

Apr

May

JunJulAug

SepOct

Nov

Dec

Jan

FebMar

Apr

May

JunJulAug

Sep

Oct

Nov

Dec

2007 2008 2009

Fene

prod

uction

But we’ve made rapid progress

Fuels strain improvement since program start

Artemisinin base strain(modified for fuel synthesis)

14

Apr

May

JunJulAug

SepOct

Nov

Dec

JanFebMar

Apr

May

JunJulAug

SepOct

Nov

Dec

Jan

FebMar

Apr

May

JunJulAug

Sep

Oct

Nov

Dec

2007 2008 2009

Fene

prod

uction

How we got there

+ process development

+ rational engineering

+ breeding

mutagenesisArtemisinin base strain(modified for fuel synthesis)

15

What can we learn from the mutants?

Illumina paired-end sequencing performed by Prognosys, Sequence assembly & analysis by Amyris

16

Mutant family tree and performance gains

M

N O

A B

C

D

E

IH

G

LF

P

K

J

Postchild

29 18

8

16

3.1

20

20

11

13

-2

7

18

0high

medium

low

No change

Causal mutations found in•Post-translational regulation•Cofactor synthesis

Most genes we’d never considered. None “on pathway”

One we had tried rationally but not with the right mutation

improvement

MutantStrain X

X

17

What about rational engineering?

Are electronics and machines the right paradigm?

Synthetic Biology: the dream of plug and play biology

18

The neutral chassis hypothesis

Add a little synthetic biology

19

Biology is designed by natural selection

It works, but it’s not always pretty

20

Too many parts kinda complexity

21

Promoter strength varies depending on its insertion siteG

ene

Exp

ress

ion

1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 41 2 3

A B C D EPromoter

Genome locus

22

How much does diversity influence pathway production?

Sporulate(haploidize)

4 diverse hybrid haploids.Select for production

FarnesenePathway

23

Impact of diversity on Mevalonate production

Ref

eren

cest

rainFo

ld in

crea

se in

mev

alon

ate

titer

ove

r ref

eren

ce

24

Frequencies for top and bottom Mevalonate pools

25

A practical approach

Enzyme kinetics

Doable but hard

Context-effects & post-transcriptional regulation

Shooting in the dark

Stoichiometry & mRNA expression

Routine

We always start here

1 Rational Semi-rationalRandom

Random

Pathway PoC Pathway Optimization

We have targeted activities here when bottlenecks become clear

2 3

This is where most of the strain improvement “action” is.

26

1. Rational: Modeling / Isotopomers

Input-Public models-Yeastcyc-Amyris knowledge-Experimental data

Output-Balanced models-Simulatable models-Matlab, Excel, GAMS format

Yeast metabolic database

27

FermentationDownstreamProcessing(DSP)

Scale-Up

2. & 3. Industrialize strain improvement

AnalyticsKnowledgeManagement

ScreeningRational StrainDesign

RandomMutagenesis

StrainEngineering

Capacity:Screen >70,000 strains / weekTest >40 2L fermentations / week

28

Continuous process improvement is critical

0.0%

1.0%

2.0%

3.0%

4.0%

5.0%

6.0%

7.0%

8.0%

5 10 15 20 25 30

Yield of parent strain

Relative improvement for a constant absolute yield gain

CV required to detect winning mutant

Assuming constant absolute yield gain per improved mutant strain. • S/N will drop as yield increases. • So too must CV.

29

The value of process control

Original strain screening assay

New assay

Relative production by shake plate assay

Relative production by shake plate assay

Yiel

d in

2L

tank

sYi

eld

in 2

L ta

nks

The reward:

Overall screening process CV <4%

Enables detection of 4% improvements w/ 5% FN and 5% FPs through 2 tiered screen

30

HT screening pipelineProcess improvement is easier said than done

31

Diagnosing sources of variation

32

Y4921

Cou

ntBetter decisions via informatics integrationLIMS systems is identifying and eliminating sources of error

Systematic drops in median plate titer traced to worn posts in one plate shaker

Stra

in s

core

33

0.940.260.00

Multivariate optimization – picking winnersInformatics integration is critical to good decisions (get data out of silos)

winner

Stress resistance

34

Strain hit from Screening(HTS Proj.)

Calculate Yield

(PV Proj.)

Tank Testing (PV Proj.)

Plate, pick, assay (HTS proj.)

MAD DB

Plate re-testing (MAD Proj.)

PV DB

HTS DB

Data Warehouse

Calculator App Filter, calculate,

store, visualize (K2Y proj.)

Informatics integration enables assessment of stress resistance

Stress resistance metric

35

Let the data guideUse empirical data mining to guide library construction, screening conditions, process dev.

36

Conclusions

• Rational engineering gets the ball rolling

• Industrialization enables rapid strain optimization– Harnessing nature’s way of “thinking”: randomness and diversity– We are doing in 4 years what used to take 12

• Continuous process improvement is critical to the success of an industrial platform– Informatics is fundamental– Data mining & omics is fundamental

Thanks to the >200 folks in R&D contributing to our success