1 Programmable Self-Assembled DNA-Based Autonomous Molecular Devices John Reif Duke University DNA Nanostructure Group John H Reif & Thomas H. LaBean Graduate Students: Harish Chandran and Nikhil Gopalkrishnan Recent Graduated Phds: Urmi Majumder, Sudheer Sahu, & Peng Yin

2 Self-Assembly in Nature Spontaneous organization of components into stable superstructures due to local interactions

3 Key to DNA Self-Assembly TTGTTTAACCT AA C AAA TTGGA 5’3’ 5’3’ Hybridization TTGTTTAACCT AA C AAA TTGGA 5’3’ 5’3’

4 Hybridization for superstructures (Mao et al: Nature00)NYU&Duke Univ (Park, et al 05) Duke Univ (Yan et al Nature03) Duke Univ (He et al 05) 2D Periodic Grid Lattices 3D Cube (Chen and Seeman, 91) (Rothemund et al 04) 1D Algorithmic Assembly (Rothemund 06) Origami - 2D Addressable Lattices cool Base Pairing cool sticky end (Yan et al: PNAS 03) Duke Univ Barcode Barcode patterning 2D Algorithmic Assembly 2D Hierarchal Assembled Lattices (Park et al: Angewandte Chemie06) Duke Univ (Lui et al PNAS 04) Duke (Yin et al Science 08) Duke&Caltech Tube Lattices

5 Compact, Complex, Robust, Flexible, Scalable, Easily Characterized Computing Device Activatable Tiles (Compact, Robust) Error Minimization Redundant Tile Design Binary Counter (Compact, Robust) Part I Stochastic Model Yield & Convergence Rates (Easily Characterized) DNA Walkers Walking on 1D & 2D Lattices (Programmable) Double-decker tiles Tiling in 3D (Scalable) Applications Reaction Catalyzation DNAzyme DNADoctor Isothermal DNA or RNA Detection

6 Double-decker tiles: Route to Assembly in 3D No tile rigid enough to create 3D periodic lattices Difficult to characterize Challenges Design a motif that can tile in 2D as well as 3D Goal Protein Crystallization: original goal of DNA nanotechnology Molecular sieve, 3D computing, host guest molecules Motivation Double-decker tiles: Route to Assembly in 3D Urmi Majumder, Abhijit Rangnekar, Thomas H. LaBean and John H. Reif in preparation

7 Cross tiles: Grid Assembly in 2D Cross Tile Symmetric Tile Figures adopted from He et al, 2005 Branched Junction Corrugation creates enormous lattices

8 Double-decker tiles: Route to Assembly in 3D 4 identical arms sticky ends 2 cross tiles held together by branched junctions Branched Junction

9 Double-decker tiles: Route to Assembly in 3D Corrugation cancels curvature of lattice => creates enormous lattices 2D Corrugation 2D Pad Programming of Double-Decker Tiles

10 Double-decker tiles: Route to Assembly in 3D 2D Lattices Yeilds: Extremely Large, Regular 2D Grids with Predominant Unidirectional Banding 10 um 2D Programmed Double-Decker Tiles

11 Double-decker tiles: Route to Assembly in 3D 3D Programming of Double-Decker Tiles 3D Generalized Corrugation cancels curvature of lattice in all 3 dimensions !

12 Summary of Results Double-decker tiles: Route to Assembly in 3D DNA Design of new motif (Double-decker tile) Flexible sticky end programming Sticky Ends can be programmed to form 2D lattices Sticky Ends can be programmed to form 3D lattices Agarose gel verification of tile formation Programming of sticky ends for 2D Lattices with corrugation AFM verification of formation of big, rigid lattices (10s of um) Fluorescence verification of formation of enormous lattices (100s of um) Analyze unidirectional banding in 2D lattices Reprogramming of sticky ends for 2D lattices without corrugation Fluorescence verification of formation of enormous lattices (100s of um) Double-decker tiles: Route to Assembly in 3D Urmi Majumder, Abhijit Rangnekar, Thomas H. LaBean and John H. Reif in preparation

13 Compact, Complex, Robust, Flexible, Scalable, Easily Characterized Computing Device Activatable Tiles (Compact, Robust) Error Minimization Redundant Tile Design Binary Counter (Compact, Robust) Part II Stochastic Model Yield & Convergence Rates (Easily Characterized) DNA Walkers Walking on 1D & 2D Lattices (Programmable) Double-decker tiles Tiling in 3D (Scalable) Applications Reaction Catalyzation DNAzyme DNA Doctor Isothermal DNA or RNA Detection

14 Stochastic Analysis of Reversible Assembly Every chemical reaction is reversible Reversible Assembly close to reality Information about time complexity, assembly yields Motivation Existing abstract model: irreversible and assumes error-free growth Kinetic tiling assembly modeled errors for DNA No framework for studying convergence rates General model for reversible assembly in 2D: hard whether infinite tiles form in the percolation problem not known in general case Challenges Goal Make simplifying assumptions Study existence of equilibrium in 2D and 3D assembly Characterize equilibrium Calculate time to equilibrium Stochastic Assembly of Self-Assembly Processes Urmi Majumder, Sudheer Sahu and John H. Reif Comp. & Theo. Nano., 5, , 2008

15 Tiling Assembly N W T E S Input1 Input 2 Output 2 Output 1 Encode computation as tiles Temperature = 2 Tiling Assembly is Turing Universal Assembly Rule: Glue type as well as glue strength have to match for assembly A tile can attach to an assembly iff the combined strength of the “matchings glues” is greater than or equal to the temperature. x y y x Counter Encoding ⋀ ⊕ y x Computational Tiles strength =1 seed y input x input strength =

16 Tiling Assembly Kinetic Model for Errors rfrf r b,2 rfrf rfrf r b, Error due to pad mismatch!

17 Stochastic Analysis of Reversible Assembly Solve important subclass of 2D assemblies Allow only monomer addition (No super-tile assemblies allowed) Pre-assembled boundary Same on/off rate for each binding or dissociation event for all tile types Binding Rule: A tile can bind to a site where it has at least two neighbors Dissociation Rule: A tile can only dissociate from a growth site where it has at most two neighbors Binding or dissociation event on one pad of a tile is independent of what’s happening on the remaining three pads Model Assumptions

18 Stochastic Analysis of Reversible Assembly Equilibrium Characterization n x n completely addressable square Let a ij denote the fraction of a tile T ij when it is free at top /right Assume σ = on probability and τ = off probability Dropping subscripts, let a’ be the next time step value of a. Then At steady state Off event On event Time Convergence: Multiplicative (<1) decrease in each time step Δ(t), distance from equilibrium decays exponentially in t

19 Stochastic Analysis of Reversible Assembly Summary of Results General characterization of equilibrium for 2D assembly Yields & Polytime Convergence to Equilibrium Completely addressable square in 2D and 3D Periodic Assembles Algorithmic Assemblies (Distribution of error at near-equilibrium) Assemblies with Partial Mismatches Correlation between Rapidly Mixing Markov Chains and Self- Assembly Stochastic Assembly of Self-Assembly Processes Urmi Majumder, Sudheer Sahu and John H. Reif Comp. & Theo. Nano., 5, , 2008

20 Compact, Complex, Robust, Flexible, Scalable, Easily Characterized Computing Device Activatable Tiles (Compact, Robust) Part III Stochastic Model Yield & Convergence Rates (Easily Characterized) Error Minimization Redundant Tile Design Binary Counter (Compact, Robust) DNA Walkers Walking on 1D & 2D Lattices (Programmable) Double-decker tiles Tiling in 3D (Scalable) Applications Reaction Catalyzation Enzyme Free DNADoctor Isothermal DNA or RNA Detection

Computational tiles Frame tiles Seed tile Error! Computational Errors (Winfree)

Error Resilience: Previous Approaches ➡ Optimizing physical conditions ‣ Decrease concentration and increase binding strength [Winfree 98] ‣ Shortcoming: Reduces speed ➡ Biochemistry Techniques ‣ Strand invasion [Chen et al 04] ‣ Shortcoming: Increase in tile set size ➡ Coding Theory Methods ‣ Proofreading Tiles, Snake Tiles, Zig-zag Tiles [Winfree et al, 2003, Chen et al 2004, Schulman et al 2005] ‣ Shortcoming: Increase in tile set size ‣ Compact Redundancy techniques [Reif et al 2004, Sahu et al 2006] ‣ Shortcoming: Ignores nucleation errors

Compact Error Correction of Computational Lattices (Reif, et al 2004) Initial Computational Tiles: Error Resilient DNA Tiles: Self-Propagation of Error Detection Makes Erroronious Assembly Unstable

24 Error Minimization in Tiling Assembly: in vitro Motivation Natural DNA self-assembly has powerful physical mechanisms for error correction & repair Artificial self-assembly needs similar mechanisms Very difficult to build large structures w/o these capabilities Challenge Minimize errors at the same scale as original assembly w/o modifying tile structure Goal Control Physical parameters to reduce errors Annealing Temperature Relative Stoichiometry of tiles Perform self-assembly w/o a scaffold Error Minimization through Optimization of Physical Parameters: Assembly of a Binary Counting Lattice using DNA Cross- Tiles, Thomas H. LaBean, Sung Ha Park, Urmi Majumder, Masahito Yamamoto, and John H. Reif, in submission (2009).

25 Characteristics of the experiment No nucleating structure used Result comparable to previous demonstration of Binary Counter (Barish et al, 2005) Error Minimization Second step annealing temperature tuned based on melting data of tiles forming grids and ribbons Relative stoichiometry of tiles tuned based on a fixed size binary counting pattern Use of a pre-assembled nucleating structure Minimize spontaneous nucleation Information about which lattices to analyze under AFM Summary of Results Error Minimization in Tiling Assembly: in vitro Error Minimization through Optimization of Physical Parameters: Assembly of a Binary Counting Lattice using DNA Cross-Tiles Thomas H. LaBean, Sung Ha Park, Urmi Majumder, Masahito Yamamoto, and John H. Reif Manuscript

26 Error Minimization in Tiling Assembly: in vitro Temperature Control After: Counting! Before: Single tile association BC2 BC3

27 Error Minimization in Tiling Assembly: in vitro Stoichiometry Control Before After: 70% reduction in Error Tile#Ratio BC12010 BC24020 BC32211 BC4189

28 Error Minimization in Tiling Assembly: in vitro Use of pre-assembled nucleating structure Minimize spontaneous nucleation Information about which lattices to analyze under AFM

29 Compact, Complex, Robust, Flexible, Scalable, Easily Characterized Computing Device Activatable Tiles (Compact, Robust) Part IV Stochastic Model Yield & Convergence Rates (Easily Characterized) Error Minimization Redundant Tile Design Binary Counter (Compact, Robust) Double-decker tiles Tiling in 3D (Scalable) DNA Walkers Walking on 1D & 2D Lattices (Programmable) Applications Reaction Catalyzation DNAzyme DNADoctor Isothermal DNA or RNA Detection

30 Error Minimization in Tiling Assembly: in silico Types of Error Mismatch Error Model assumes directional growth (i/p to o/p) Model assumes T=2 rule (at least two correct binding required) Also known as error by insufficient attachment Spontaneous Nucleation Error Assembly in absence of seed Challenge Minimize errors At the same scale as original assembly Use already existing DNA nanostructures with minimal modifications Handle all kinds of errors (related to the tile assembly model) Goals Enforce model assumptions at the same scale as original assembly

Activatable Tiles: Basic Idea ➡ Tiles are initially inactive ‣ o/p pads protected and not available for hybridization ➡ Tiles transition to active state and o/p pads are exposed only when the correct neighbors bind to its input pads

32 Error Minimization in Tiling Assembly: in silico Error by insufficient attachment (T=2) Activatable Tiles: Working Principle ➡ Tiles are initially inactive ‣ o/p pads protected and not available for hybridization ➡ Tiles transition to active state and o/p pads are exposed only when the correct neighbors bind to its input pads

33 Activatable Tile Correct Growth Error Minimization in Tiling Assembly: in silico One correct i/p match induces the other i/p deprotection

34 Activatable Tile prevents errors by insufficient attachment Second i/p is not deprotected Error Minimization in Tiling Assembly: in silico

35 Error Minimization in Tiling Assembly: in silico Small probability of error from the tiles that leave a growth site after being completely deprotected. Input deprotection reversible Output deprotection irreversible Source of Error

DNA Implementation ➡ Strand Displacement for Input Deprotection ➡ DNA polymerization for Output Deprotection ‣ Particularly Effective over long distances (e.g. tile cores)

37 Strand Displacement DNA Polymerization Strand displacing DNA Polymerization DNA Strand Displacement Using Polymerase Phi 29 for Strand Displacement: - Replicative polymerase from bacteriophage Phi29 - Phi29 polymerase can travel at the rate of 2000 nucleotides per minute at room temperature - This polymerase has exceptional strand displacement and processive synthesis properties

Activatable Tiles: Basic Idea in 1D

DNA Design of 1D Activatable Tile

A Reaction Pathway Stage 0 S2S1AS1BH S1A’S1B’ P’M S2’ 5’ 3’ 5’ Tile Core E Tile 1(Protected) 5’ 3’ Tile 2(Unprotected) H’S1A’S1B’ S3 Stage 2 Hybridization of sticky ends by displacement of the protection strand S2 S1AS1BH S1A’S1B’P’ M S2’ 5’ 3’ Tile Core E 3’ H’ S1A’S1B’ S3 Stage 1 S2S1AS1B H S1A’ S1B’ P’M S2’ 5’ 3’ Tile Core E H’ S1A’ S1B’ 3’ S3 Toehold hybridization 5’

Polymerase Stage 5 S2S1A S1B H S1A’ S1B’ P’ M S2’ 5’ 3’ Tile Core E 5’ 3’ H’S1A’S1B’S3 Complete polymerization of the primer and dehybridization of protection strand from the output sticky end S1AM’ Exposed output sticky end 5’ P 3’ P S2S1A S1B H S1A’S1B’ P’ M S2’ 5’ 3’ Tile Core E 5’ 3’ H’ S1A’S1B’ S3 Primer binding to now available template (protection strand) Stage 3 5’ 3’P Primer

42 Error Minimization in Tiling Assembly: in silico GS Original ∝ E 2 GS: Growth Speed E: Error Rate GS 2x2 ∝ E GS Activatable GS Original > e - ∊ Gse E Activatable E Original = e - ɣ Gse 0< ɣ << ∊ <1

43 New kind of tile : Activatable tile Tile set size same as before Basic nanostructure: existing tile types Errors handled Minimizes error due to insufficient attachment (proof) Minimizes nucleation error Allows self-healing (proof) Summary of Results Error Minimization in Tiling Assembly: in silico Protection / deprotetcion mechanism through strand displacement + polymerization DNA Design of 1D/2D activatable tile system Applications beyond computing Concentration System Reaction Catalyzation Activatable Tiles for Compact, Robust Programmable Assembly and other Applications Urmi Majumder, Thomas H. LaBean, and John H. Reif DNA 13, LNCS 4848, 15-25, 2007

Summary ➡ Activatable tiles reduce error in assembly by virtue of physical design of the tiles (use of DNA strand displacement and DNA polymerization) ➡ Other Potential Applications: ‣ A Chemical Concentration Probing System ‣ Chemical Reaction Catalytic System ➡ Current Work: ‣ Test a 1D Deprotection System ➡ Open Question: ‣ Overlay Redundancy Technique+ Activatable Tiles

45 Compact, Complex, Robust, Flexible, Scalable, Easily Characterized Computing Device Part VI Stochastic Model Yield & Convergence Rates (Easily Characterized) Error Minimization Redundant Tile Design Binary Counter (Compact, Robust) Activatable Tiles (Compact, Robust) in silico DNA Walkers Walking on 1D & 2D Lattices (Programmable) Double-decker tiles Tiling in 3D (Scalable) Applications Reaction Catalyzation DNAzyme DNADoctor Isothermal DNA or RNA Detection

DNA Walker Devices: Formulation & First Designs [Reif, 2002] Designs for the first autonomous DNA nanomechanical devices that execute cycles of motion without external environmental changes. Walking DNA device Rolling DNA device Use ATP consumption Use hybridization energy These DNA devices translate across a circular strand of ssDNA and rotate simultaneously. Generate random bidirectional movements that acquire after n steps an expected translational deviation of O(n 1/2 ).

First Autonomous DNA Walker 2004: Peng Yin, Hao Yan, Xiaoju G. Daniel, Andrew J. Turberfield, John H. Reif, A Unidirectional DNA Walker Moving Autonomously Along a Linear Track, Angewandte Chemie Volume 43, Number 37, Sept. 20, 2004, pp B C D A Track Anchorage A Walker * Ligase PflM I BstAP I Restriction enzymes

53 Autonomous Motion of the Walker

autonomous, programmable, and further require no protein enzymes. ________________________ The basic principle involved is inspired by a simple but ingenious molecular device due to Mao et al – Mao used DNAzyme to traverse on a DNA nanostructure, but was not programmable (it did not executed computations). Programmable Autonomous DNA Nanorobotic Devices Using DNAzymes John H. Reif and Sudheer Sahu

Our DNAzyme based designs 1. DNAzyme calculator : a limited ability computational device 2. DNAzyme FSA: a finite state automata device, that executes finite state transitions using DNAzymes –extensions to probabilistic automata and non-deterministic automata, 3. DNAzyme router: for programmable routing of nanostructures on a 2D DNA addressable lattice 4. DNAzyme porter: for loading and unloading of transported nano-particles 5. DNAzyme doctor : a medical-related application to provide transduction of nucleic acid expression. –can be programmed to respond to the under-expression or over-expression of various strands of RNA, with a response by release of an RNA –operates without use of any protein enzymes.

DNAzyme FSA (inputs, transitions)

DNAzyme Crawler

DNAzyme Calculator

54 A “Molecular-Racecar” on Circular Track Used Power of Strand-displacing Polymerase Used Polymerase Phi 29 to push wheel W on circular track T -Protector BQ prevents W from moving on its own -Powerful strand displacement capability of Phi 29 during polymerization dislodges BQ from track => much faster & forceful movement than other DNA Walkers Sudheer Sahu, Thom H. LaBean, John H. Reif, A DNA Nanotransport Device Powered by Polymerase Phi29, Nanoletters, 2008 A DNA Nanotransport Device Powered by Polymerase Phi29 Polymerase Phi 29 Replicative polymerase from bacteriophage Phi29 Phi29 polymerase can travel at the rate of 2000 nucleotides per minute at room temperature This polymerase has exceptional strand displacement and processive synthesis properties -Experimental Demonstrations via FRET and Gel data

55 Compact, Complex, Robust, Flexible, Scalable, Easily Characterized Computing Device Summary Stochastic Model Yield & Convergence Rates (Easily Characterized) Error Minimization Redundant Tile Design Binary Counter (Compact, Robust) Activatable Tiles (Compact, Robust) Double-decker tiles Tiling in 3D (Scalable) DNA Walkers Walking on 1D & 2D Lattices (Programmable) Applications Reaction Catalyzation DNAzyme DNADoctor Isothermal DNA or RNA Detection

Other Applications of Activatable Tiles Molecular Sensing and Concentration System

Other Applications of Activatable Tiles Reaction Catalyzation

Other Applications of Activatable Tiles Reaction Catalyzation

DNAzyme Doctor (state diagram)

DNAzyme Doctor(RNA expression)

61 Application of DNA Nanotechnology: Isothermal DNA and RNA Detection Isothermal detection protocols: exquisitely sensitive detection of specific DNA or RNA sequences. Report target DNA detection via nanoparticle colorimetric detection. Two Isothermal Detection Techniques Demonstrated: (1) Superlinear Hybridization Chain Reaction (HCR) - Based on linear hybridization chain reaction of (Dirks04). -triggered by target DNA or RNA - New DNA detection protocol using a superlinear hybridization cascade reaction - No use of Enzymes - Superlinear Detection Response - nanoparticle-based colorimetric readout. (2) Cross-Catalytic Deoxyribozymogen Reaction (DRZ) - No use of Protein Enzymes - Based on cross-catalytic DNAzyme-based reaction of (Levy03) -triggered by target DNA or RNA - Exponential Detection Response - nanoparticle-based colorimetric readout. Diagnostic Applications: Detectiion of Disease Sequences (DETECTIONChlamydia & HIV) The modification of the HCR and DRZ methods to detect target sequences in Chlamydia trachomatis bacterial DNA & HIV viral RNA/DNA sequences. 61

62 Compact, Complex, Robust, Flexible, Scalable, Easily Characterized Computing Device Summary Stochastic Model Yield & Convergence Rates (Easily Characterized) Error Minimization Redundant Tile Design Binary Counter (Compact, Robust) Activatable Tiles (Compact, Robust) Applications Reaction Catalyzation Enzyme Free DNADoctor Isothermal DNA or RNA Detection Double-decker tiles Tiling in 3D (Scalable) DNA Walkers Walking on 1D & 2D Lattices (Programmable)