Computing with Life: Digitally Programmed Cells

1
Computing with LifeDigitally Programmed Cells

• Tom Knight
• MIT Artificial Intelligence Laboratory

2
1 micron ---------------------------
3
A New Engineering Discipline

• Living Systems are Special
• They self replicate
• They are self powered from common chemicals
• small
• They clean up after themselves
• They provide an interface to the chemical world
• precision chemical construction
• antibody sensors and catalytic effectors
• They Compute
• But we need to engineer and control their
  computation
• We need to standardize their I/O interfaces
• They cooperate to fabricate complex structure
• Information Rich Materials

4
Implementing the Digital Abstraction

• In-vivo digital circuits
• signal concentration of specific protein
• computation regulated protein synthesis decay
• The basic computational element is an inverter
• Logical operation
• More importantly signal restoration
• The output must be better than the input
• low gain in the high and low states
• high gain in intermediate states
• nonlinear transfer curve

5
Biochemical Inverters
6
Digital Circuits

• Combining properly designed inverters, any
  digital circuit can be built

C

A
C
D
D
gene
B
C
B
gene
gene

• proteins are the wires, genes are the gates
• NAND gate wire-OR of two genes

7
Proof of Concept Circuits

• They work in simulation
• They work in vivo
• Models poorly track their behavior

RS-Latch (flip-flop)
Ring oscillator
_ R
A
_ R
_ S
A
B
time (x100 sec)
B
B
_ S
C
A
time (x100 sec)
time (x100 sec)
Gardner Collins 2000
Elowitz Leibler 2000
8
Bio Circuit Design (TTL Data Book)

• Data sheets for components
• imitate existing silicon logic gates
• new primitives from cellular regulatory elements
• e.g. an inverter that can be induced
• Assembling a large library of components
• modifications that yield desired behaviors
• Constructing complex circuits
• matching gates is hard
• need standard interfaces for parts
• from black magic to you can do it too

9
Biological InverterTransfer Curves

• Dramatic differences
• Lambda cI vs. lacI
• Difficult measurements
• Single cell expression necessary
• Cell type and cell state has a huge effect
• Dual labelling with CFP / GFP / YFP
• Interchange approach to measurement
• Techniques out of VLSI design
• dual rail signalling
• self biased circuits
• Engineered binding sites and proteins

10
Naturally Occurring Sensor and Actuator Parts
Catalog

• Sensors
• Light (various wavelengths)
• Magnetic and electric fields
• pH
• Molecules
• Autoinducers
• H2S
• maltose / arabinose / lactose
• serine
• ribose
• cAMP
• NO
• Internal State
• Cell Cycle
• Heat Shock
• Chemical and ionic membrane potentials

• Actuators
• Motors
• Flagellar
• Gliding motion
• Light (various wavelengths)
• Fluorescence
• Autoinducers (intercellular communications)
• Sporulation
• Cell Cycle control
• Membrane transport
• Exported protein product (enzymes)
• Exported small molecules
• Cell pressure / osmolarity
• Cell death

11
Engineering the Lux Operon

• Starting point
• Vibrio fischeri strain MJ-1
• infects the light organ of Monocentris japonicus
  Japanese Pinecone fish
• Also free ocean swimming
• Dark when free swimming
• Emits light when in the light organ
• Light production cascade
• Autoinducers
• small communications molecules (homo-serine
  lactones)
• both emitters and detectors

12
Cloning the lux Operon into E. coli

• First, we shotgun cloned the lux Operon from
  Vibrio fischeri to form plasmid pTK1
• Expressed in E. coli DH5a ? showed
  bioluminescence
• Sequenced the operon Genbank entry AF170104

13
Split -- engineer componentsand interfaces

• Autoinducer fabrication (LuxI)
• Autoinducer response (Pr promoter)
• Luciferase genes (Lux A, B)
• Better ones from Xenorhabdus luminescens
• Aldehyde production (Lux C, D, E, G)
• Bidirectional terminator

14
Experiment III Controlled Sender

• Genetic networks for controlled sender
  receiver

VAI
VAI

E. coli strain expresses TetR (not shown)

• Logic circuit diagrams for controlled sender
  receiver

LuxR
GFP
TetR
VAI
VAI
aTc
aTc
pLuxI-Tet-8
pRCV-3
15
Experiment III Controlling Sender

• Figure shows ability to induce stronger signals
  with aTc
• Non-induced sender (pLux8-Tet-8) receiver cells
  grown seperately _at_37C to late log phase
• Cells were combined in FL600, and sender cells
  were induced with aTc
• Data shows max fluorescence after 4 hours _at_37 C
  for 5 separate cultures plus control positive
  cultures have same DNA ? variance due to OD

positive control
negative control
16
Engineered Minimal Organisms

• simple
• understood
• malleable
• controlled
• Start with a simple existing organism
• Remove structure until failure
• Rationalize the infrastructure
• Learn new biology along the way

The chassis and power supply for our computing
17
Hutchison experiment

• Randomly insert transposons into locations in the
  Mycoplasma genitalium genome
• Select for cells with an insert
• Cells which grow survived the destruction of the
  interrupted gene
• Sequence from the transposon, report which genes
  were not needed
• Only 293 out of 453 genes were essential

Nature, Dec. 1999
18
Plan

• Locate a non pathogenic simple cell
• plant and insect mycoplasma species
• Mesoplasma florum, M. lactucae, M. entomophilum
• Sequence
• Develop plasmids
• Eliminate restriction systems
• Locate and remove unnecessary genes
• Rationalize promoters, codon usage
• Reclaim amino acid coding space
• Understand the cell metabolism and control
• Add debugging and control hooks

19
Recent progress

• Chose Mesoplasma florum
• Easily grown -- saturation in 36 hours
• Sequenced 3 to know what we are up against
• Genome is approximately 870 Kb
• We are developing tools to examine protein
  expression
• We have discovered that two supposedly distinct
  species are essentially identical
• exact identity of 16S ribosomal RNA sequence
• near exact identity of protein expression
  patterns
• Engineering replicative plasmids

20
Mesoplasma florum
21
PFGE of Mesoplasmagenomic DNA
y yeast marker l lambda marker me Mesoplasma
entomophilum mf Mesoplasma florum ml Mesoplasma
lactucae
1.2 agarose 9C 6V/cm Ramped 90 - 120 sec 48
hours
22
Mesoplasma SDS-PAGE
ME Mesoplasma entomophilum MF Mesoplasma florum
ML Mesoplasma lactucae whole cell 10 Tris-HCl
gel
23
I think in order to get to a petaflop we have to
somehow reduce the size of our components from
the micron size to the nanometer size. But there
are lots of interesting things happening at the
nanometer scale. During the past year I have
read articles which make my jaw drop. They
arent from our community. They are from the
molecular biology community, and I can imagine
two ways of riding on the coattails of a much
bigger revolution than we have. One way to do
that is to make computing devices out of
biological elements -- but Im not comfortable
about that because I am one and I feel
threatened. We can also use biological tools to
manufacture non-biological devices to
manufacture the things that are more familiar to
us and are more stable, in the sense that we
understand them better
Seymour Cray, January 1994, Petaflops workshop
24
Molecular Electronics

• Semiconductor technology ends lt .05 microns
• Statistical placement of atoms
• poor matching between devices
• Alternative is precise atomic placement and
  identical devices
• Path to continued performance enhancement of
  electronics and extension of Moores Law
• Every reason to believe that this technology will
  be as important in the next century as silicon is
  today.

25
Molecular Fabrication

• Nanoscale molecular electronics requires control
  over highly structured information-rich placement
  of individual atoms
• Biochemistry is our best hope of achieving this
  control
• We can isolate structure from function
• an information rich substrate, biochemically
  computed
• Collagen?
• a binding mechanism (antibodies?)
• high performance molecular devices
• carbon nanotubes
• conjugated polymers (Carotenes)