Sequential self-assembly of DNA functionalized droplets

Complex structures and devices, both natural and manmade, are often constructed sequentially. From crystallization to embryogenesis, a nucleus or seed is formed and built upon. Sequential assembly allows for initiation, signaling, and logical programming, which are necessary for making enclosed, hierarchical structures. Although biology relies on such schemes, they have not been available in materials science. Here, we demonstrate programmed sequential self-assembly of DNA functionalized emulsions. The droplets are initially inert because the grafted DNA strands are pre-hybridized in pairs. Active strands on initiator droplets then displace one of the paired strands and thus release its complement, which in turn activates the next droplet in the sequence, akin to living polymerization. Our strategy provides time and logic control during the self-assembly process, and offers a new perspective on the synthesis of materials.


Supplementary Figures
Supplementary Figure 1 | A Native PAGE shows negative controls of other unintended "cross-talk" reactions, leaving the designed sequential assembly the only pathway. Column Ld is the 10 bp ladder. Column S 0 serves as a reference for single strands S 2 , S 4 , and S 6 with the similar molecular weights. S 7 marks the terminator hairpin. L 3 is a reference for L 1 and L 2 (partially hybridized strands). The results of column S 0 + L 2 to column S 6 + L 2 confirm that all the unintended cross-talk reactions won't happen. . The three curves plotted in the figure above are emission fluorescence intensities with the background subtracted. The peak values are labeled and used to calculate the amount of streptavidin bound to emulsion droplets. The inset plots the amount of streptavidin dosed initially versus the peak intensity of the subnatant. We linear fit the data and found the interception of y-axis to be ~6.6 pmol, which is the amount of streptavidin consumed by 5 µl of creamed biotinylated emulsion droplets.

Supplementary Methods
Measurement of DNA strands density on droplet. A fixed amount of emulsion droplets (functionalized with biotinylated phospholipids) was mixed with varying amount of streptavidin and then incubated. After centrifugation the subnatants from each sample containing unbound streptavidin were placed in a fluorescent spectrometer (Horiba PTI QuantaMaster 400) for fluorescence measurements. The signal intensities detected from different samples were used to calculate the amount of streptavidin bound to emulsion droplets ( Supplementary Fig. 2). We first diluted 20 µl of creamed emulsion (4.7-µm droplets) in 600 µl TMS buffer, and split it into four parts (labeled with sample A, B, C and D) evenly. 2 µl, 1 µl, 0.5 µl and 0 µl (control group) of Alexa® Fluor 488 streptavidin (37.88 µM) were added to sample A, B, C and D respectively. The samples were then gently tumbled at room temperature for 0.5 hr. After incubation, the samples were centrifuged to cream the droplets. From each sample, 5 µl of the subnatant was taken out and diluted to 800 µl with the same buffer for fluorescence measurements. The curve from the control group (sample D) was set as the background. The three curves plotted in the figure above were emission fluorescence intensities with the background subtracted. The peak value of each curve should be proportional to the amount of unbound streptavidin. It can be written as I peak ∝ n subnatant = n tot -n emulsion , where n subnatant , n tot and n emulsion refer to the amount of streptavidin in the subnatant, the amount dosed at the beginning, and the amount bound to emulsion, respectively. Plotting n tot against I peak , we then linear fit the data and found n emulsion from the interception of y-axis to be 6.6 pmol. Knowing that the volume fraction of a random close packing of droplets is ~0.6 and the average radius is 2.35 µm, we can calculate the number of droplets in 5 µl of creamed emulsion. The coverage of streptavidin molecules on the droplet surface is then obtained as ~1000 /µm 2 . This is also used as the surface coverage of DNA strands in our paper, assuming 1: 1 binding ratio of streptavidin to biotinylated DNA. For the 3.4-µm droplets we assume they have similar surface coverage because they showed similar brightness from confocal fluorescent microscopy.
Chamber preparation. The flow chamber has a sandwich structure with a glass slide as the bottom layer, two small pieces of glass slides (thickness = 1 mm) as spacers and a cover slip on the top. The dimension of the chamber is 10 mm×2 mm×1 mm (~20 µl volume). The cover slip was plasma etched first and then treated with HMDS (hexamethyldisilazane, from Sigma Aldrich) overnight. Before loading sample, the chamber was soaked in a 1% pluronic F127 buffer for 2 minutes and then gently dried with filtered nitrogen gas. After the sample was loaded, the two open ends were sealed with UV curing adhesives (NOA 68, Norland Products, Inc.).

Supplementary Note 1 Kinetics model for the initiated polymerization process.
For 4.7 µm droplets, the average aggregation time for each step can be written as [1]: . (1) τ DLA can be calculated from the equation: . ( The number densities of B, C, D and E are equal, leading to an identical τ DLA for each step of binding. Using R d = 2.35 µm, C 0 = 6×10 2 mm -2 and D 0 = 0.037 µm 2 s -1 , we get τ DLA = 35 minutes (~ 0.5 hour) from the above equation.
In the experiment one part of the active DNA strands S 0 , along with four parts of neutral strands (Poly-T), are pre-coated onto the surface of droplet A. This 20% ratio is chosen to lower the branching ratio, while not greatly slowing down the binding. The measured streptavidin density on surface (equivalently the overall surface coverage of DNA strands) is ~ 1000/µm 2 . The number density of S 0 is ~ 200 µm -2 .
In comparison the active strands on droplets B, C or D are released from the adhesion patches formed from the previous binding, and is determined by the size of the patch. A previous study [2] shows that the arrangement of DNA binders within the patch can be approximated as 2D close packing of streptavidin. Thus N DNA = A patch /A stvd . From our measurements the average size of an adhesion patch is around 400 nm in diameter, equivalently A patch ~ 0.13 µm 2 . Then the surface coverage of active strands on droplet B, C or D is around 20 µm -2 . This discrepancy accounts for the difference of τ R for A -B binding and τ R for the B -C, C -D and D -E bindings.
In the model [2], τ R is the reaction time of a pair of droplets with complementary stickyend DNA on their surfaces staying within the reaction region (thickness L ~ 20 nm; area S ~ 0.1 µm 2 ). τ R is reduced by the number of possible bonding configurations N G from the equation τ R ≈ τ r /N G , where τ r is the reaction time if there is only one active strand in the reaction region. N G is the number of active strands in the reaction region, which is the product of the number density of active strand and the area of the reaction region. N G ~ 20 for A -B binding, and ~ 2 for B -C, C -D and D -E binding. τ r is estimated as In our case there is only one active strand on a droplet and many receptor strands (~ 500 strands/µm 2 for DNA loops) on the other. τ d can be calculated using smoluchowski diffusion equation: , where C DNA = 500 µm -2 , D DNA ~ 1 µm 2 s -1 [2] and !"# ! is the area that one strand occupies (~ 20 nm 2 ). Putting these parameters into the above equation, τ d is ~ 100 µs. τ h is the hybridization time for a pair of complementary DNA strands on the opposing droplets 8π D DNA C DNA to bind when the droplets are held within an interaction distance, and τ t is the translational transit time for a pair of complementary DNA strands to stay in their interaction zone. Previous study [1] has shown τ h is on the order of 1 s to 10 s, and τ t is approximately 20 µs. With all the numbers plugged in, τ r is 5 s to 50 s, and τ R is 0.25 s to 2.5 s for A -B binding, and 2.5 s to 25 s for B -C, C -D and D -E binding. The second term on the right hand side of Equation (1) ranges from 0.2 hours to 2 hours for A -B binding, and 2 hours to 20 hours for B -C, C -D and D -E binding. Then the average aggregation time τ ranges from 1.2 hours to 3 hours for A -B binding, and 3 -21 hours for B -C, C -D and D -E binding.
We also repeated the experiment using smaller droplets (diameter ~ 3.4 µm) to see if the kinetics of the polymerization process will change. Below is the calculation of the above model using the parameters for the 3.4 µm droplets.
Using R d = 1.7 µm, C 0 = 12×10 2 mm -2 and D 0 = 0.08 µm 2 s -1 , we first get τ DLA = 8 minutes from supplementary equation (2). The coverage of the initiator strands S 0 on droplet A is 10%. The average size of the patch shrinks due to the decrease of the droplet size. Here we simply assume at equilibrium the enthalpy gain from DNA hybridization (∝ the number of bonds, which is proportional to A patch ) balances the energy burden from the deformation (∝(A patch /A droplet ) 2 ). Thus A patch is proportional to the forth order of the droplet's size, which is around 0.04 µm 2 for R d = 1.7 µm. Consequently, the surface coverage of active strands on droplet B, C or D is around 12 strands/µm 2 . Also the numbers of active strands within the reaction region change to N G ~ 5 for A -B binding; and ~ 0.6 for B -C, C -D and D -E binding. τ d , τ h and τ t are the same as in the previous situation. Thus τ R ranges from 1 s to 10 s for A -B binding, and 8 s to 80 s for B -C, C -D and D -E binding. Consequently, the second term on the right hand side of Equation (1) ranges from 0.5 hours to 5 hours for A -B binding, and 4 hours to 40 hours for B -C, C -D and D -E binding. Since τ DLA is much less, the second term can be directly used as the average aggregation time in this situation.
From the calculation we can see the kinetics of the assembly is not much changed by using 3.4 µm droplets. This is also confirmed in the experiment, as shown in Supplementary Fig. 3. The reasons are as follows: (1) Smaller droplets with higher diffusion coefficient encounter more frequently (decrease τ DLA ); (2) However the area of the patch formed from each binding is much smaller due to its quadratic dependence on the droplet size, which indicates there are much less DNA strands released from each step of reaction. This increases τ R ,and as a consequence, balances the decrease of τ DLA .