Toward Memory in a DNA Brush: Site-Specific Recombination Responsive to Polymer Density, Orientation, and Conformation

Site-specific recombination is a cellular process for the integration, inversion, and excision of DNA segments that could be tailored for memory transactions in artificial cells. Here, we demonstrate the compartmentalization of cascaded gene expression reactions in a DNA brush, starting from the cell-free synthesis of a unidirectional recombinase that exchanges information between two DNA molecules, leading to gene expression turn-on/turn-off. We show that recombination yield in the DNA brush was responsive to gene composition, density, and orientation, with kinetics faster than in a homogeneous dilute bulk solution reaction. Recombination yield scaled with a power law greater than 1 with respect to the fraction of recombining DNA polymers in a dense brush. The exponent approached either 1 or 2, depending on the intermolecular distance in the brush and the position of the recombination site along the DNA contour length, suggesting that a restricted-reach effect between the recombination sites governs the recombination yield. We further demonstrate the ability to encode the DNA recombinase in the same DNA brush with its substrate constructs, enabling multiple spatially resolved orthogonal recombination transactions within a common reaction volume. Our results highlight the DNA brush as a favorable compartment to study DNA recombination, with unique properties for encoding autonomous memory transactions in DNA-based artificial cells.

#38208 (Addgene), and inserted into pIVEX to be expressed under the T7 promoter using NEBuilder HiFi DNA assembly cloning kit (New Englands Biolabs). The attP (GTCGTGGTTTGTCTGGTCAACCACCGCGGTCTCAGTGGTGTACGGTACAAACCCCGAC) and attB (TCGGCCGGCTTGTCGACGACGGCGGTCTCCGTCGTCAGGATCATCCGGGC) sequences were inserted downstream to the T7 promoter or in the 5' untranslated region (UTR) of the GFP gene and E. coli 16S ribosomal RNA (rRNA) genes according to Table S1. A RiboJ ribozyme sequence(1) that self cleaves the mRNA upon transcription was inserted between the att site and the ribosome binding site (RBS) of the GFP gene (Table S1) to ensure that the mRNA secondary structure is not inhibiting expression of the downstream gene. A Broccoli aptamer (2) was inserted in Helix 6 of the 16S rRNA for fluorescent labeling upon transcription, and an HDV ribozyme (3) was inserted at the 3' end for exact cleavage of the 16S rRNA, as described in (4).
Preparation of Linear DNA for DNA Brushes: Linear double stranded DNA was amplified from pIVEX plasmids containing the relevant constructs using PCR, with the KAPA HiFi HotStart ReadyMix PCR kit (KAPA Biosystems). One primer was modified with 5' biotin, and the other was either plain or fluorescently labelled on the 5' end with ATTO-647, ATTO-488 or Alexa Fluor-647 fluorophores (Integrated DNA technologies). For all constructs, except for one, the fluorescently labeled or plain forward primer was complementary to a region 200-300 bp upstream of the T7 promoter (or the attP sequence in promoterless constructs) and the biotinylated reverse primer complementary to a region downstream to the T7 terminator. The construct with an att site positioned 100bp from the surface (Fig. 5) was prepared with a biotinylated forward primer complementary to a region 61bp upstream to the promoter and a plain reverse primer. The DNA was incubated for 5 minutes with streptavidin (Sigma) at a 1.4 SA-DNA ratio in order to create DNA-SA conjugates, and was then diluted to a final concentration of 150nM in 1x phosphate buffered saline (PBS) supplemented with 7% glycerol, the latter added to reduce evaporation at the following DNA surface deposition step.
Photoactivable Biocompatible Monolayer Assembly: Fused-silica slides (24 x 24 x 1 mm, UQG Optics) were coated with a photoactivable biocompatible monolayer according to a previously published protocol (5). Briefly, the fused-silica slides were cleaned in boiling ethanol for 10 minutes followed by sonication for 10 minutes and base piranha cleaning (H2O2:NH3:H2O; 1:1:4, heated to 70°C for 10 minutes). The slides were then coated with a polymer composed of a polyethylene glycol backbone with a protected amine at one end, and a triethoxysilyl group at the other end. The slides were incubated with the polymer solution (0.2 mg/mL in Toluene) for 20 minutes during which the monolayer is formed, and then washed and dried. The coated slides were exposed to 365 nm UV light (2.5 J/cm2) through a custom photomask containing an array of 30 µm hexagons (CAD/Art Services) using UV-KUB (Kloe). Surface amines located inside hexagons were exposed to a saturating UV dose and were fully deprotected, whereas surface amines located between hexagons were partially deprotected due to undesired leaks through the mask.
In the transcription-based recombination assay the hexagonal pattern was not required and the surface was exposed using UV-KUB without a mask. After the UV exposure the slides were immediately covered with biotin N-hydroxysuccinimidyl ester (biotin-NHS, Pierce) dissolved in a borate buffer solution (0.5 mg/mL) and incubated for 30 minutes, during which the biotin bound covalently to the exposed amine groups. The slides were then washed and dried.  The power laws were obtained by fitting the x and y axes' logarithms to a linear function using the least squares method, and the square roots of the variance of these values are the estimated errors.

Simulation of recombination:
Recombination simulations were carried out in Python 3.
Each point of a 150x150 grid was randomly assigned to be a DNA molecule of type A, B, or noncoding, with predetermined probabilities. Each molecule of type A was randomly recombined with a molecule of type B within twice a given radius accessible by each molecule, and both of these molecules could no longer recombine. Molecules of type A and B that had no molecules of the other type within reach did not recombine.
Recombined molecule fraction was calculated as the number of recombined A and B molecules divided by the grid size.   Figure S1. Broccoli signal is proportional to gene density. Broccoli signal after three minutes of expression as a function of the DNA signal of the gene-promoter construct from the experiment shown in Fig. 3. The gene-promoter construct DNA was labeled in ATTO647 dye, and the broccoli signal measured using TIRF microscopy. The broccoli signal appears to be linear in the initial amount of promoter-gene construct.    Fig. 3. Each curve corresponds to a different gene ratio between the gene construct and the turn-off DNA: 0:1 (blue), 1:5 (orange), 1:2 (yellow), 1:1 (violet), 2:1 (light blue), 5:1 (green) and 1:0 (red). Each curve is averaged over 15 repeats. Fluorescence signal is background-subtracted and normalized by the expression in a full brush with no recombination (dark red). Figure S5. GFP turn-on assay solution experiment. All constructs as shown in Fig. 4 A were added as plasmids to a PUREfrex 2.0 solution reaction at 3nM, and the Bxb1 integrase gene at 1nM. The transcription-translation reactions were deposited in a Vbottom 96-well microplate (Costar) so that each well contained 5 µl of solution, and GFP fluorescence was measured every 3 minutes in a microplate reader (ClarioStar). Figure S6. Repeats of the kinetic data in Fig. 4 D. HA-GFP trapping kinetics independent of recombination (black) or mediated by the integrase (red) expressed from the brush along with the promoter and HA-GFP constructs, at a ratio of 2:1:1. Figure S7. A single repeat of the kinetic data in Fig. 4 E. HA-GFP trapping kinetics independent of recombination (black and purple) or mediated by the integrase (red and pink). The Bxb1 integrase was expressed from a plasmid added to the bulk at a concentration of 1 nM, and the promoter and HA-GFP constructs were expressed from the DNA brush. The data shown in Fig. 4 E is presented alongside the repeated data for comparison.