Evolutionary selection of DNA nanostructures for cellular uptake

This study introduces an evolutionary selection strategy that iteratively screens large libraries of DNA nanostructures against mammalian cells to identify and validate specific structural designs with enhanced, cell-type-specific internalization capabilities for targeted drug delivery.

The Gist

Imagine you are trying to find the perfect key to open a specific lock (a cell) to deliver a medicine inside. In the world of DNA nanotechnology, scientists have been trying to design these "keys" (nanostructures) by hand, testing them one by one. It's like trying to find a needle in a haystack by looking at one needle at a time. It's slow, and you might miss the best one.

This paper introduces a smarter, faster way: Evolutionary Selection. Instead of designing one key at a time, the researchers created a massive library of millions of different DNA shapes and let the cells themselves decide which ones are the best.

Here is how they did it, broken down into simple steps with some fun analogies:

1. Building the "DNA Lego" Library

First, the team didn't build one giant structure. Instead, they made a huge bag of small, pre-made DNA "Lego pieces" (fragments) of different shapes and sizes.

• The Trick: They designed these pieces so they could be glued together randomly to form long, single strands of DNA. Think of it like a bag of random Lego bricks that snap together to form weird, unique shapes.
• The "ID Tag": Crucially, every single structure they made had a tiny barcode (a Unique Molecular Identifier or UMI) attached to it. This is like putting a serial number on every Lego creation so they can track exactly which one is which later.

2. The "Cellular Dating" Game

Next, they threw this massive bag of DNA shapes into a petri dish with two types of cells:

• HEK293T: A human kidney cell (good at taking things in).
• RAW264.7: A mouse immune cell (very greedy at eating things).

They let the cells "date" the DNA for a few hours. The cells naturally swallow (internalize) some of these DNA shapes and spit others out.

• The Selection: The researchers washed away the DNA that stayed on the surface. They only kept the DNA that the cells actually swallowed.
• The Harvest: They broke the cells open (lysis) to get the DNA back out. Now, they had a "survivor pool" containing only the shapes that the cells liked enough to eat.

3. The "Copy and Refold" Cycle (The Evolution)

Here is where the magic happens. They took the DNA they recovered from the cells, made millions of copies of it (PCR amplification), and then folded it back into its 3D shape.

• The Loop: They put this "survivor pool" back into fresh cells for another round.
• The Result: They did this 10 times. Just like natural selection in evolution, the shapes that were best at getting into the cells got copied more and more. The shapes that were bad at getting in disappeared. By round 10, the pool was dominated by the "champions" of cellular uptake.

4. Reading the Winners

After 10 rounds, they used high-tech DNA sequencers (like a super-fast barcode scanner) to read the "serial numbers" (UMIs) and the shapes of the winning DNA.

• The Discovery: They found that certain shapes were much better at entering specific cells.
  • In the kidney cells, the winners were often specific compact shapes.
  • In the immune cells, the cells were so "promiscuous" (greedy) that they ate almost everything, making it harder to find a specific winner, but they still found some favorites.
• The Surprise: They also found that some shapes were accidentally made from human DNA fragments that got mixed in. While not the intended "design," these accidental shapes were surprisingly good at getting into cells, proving that nature (or accident) can sometimes find solutions humans miss.

5. The Final Test

To prove these winners were real, they built the top candidates from scratch (synthesized them) and tested them again.

• The Proof: Using flow cytometry (a machine that counts glowing cells) and microscopes, they confirmed that the "winning" shapes indeed entered the cells much better than random shapes.
• The Twist: When they tested these winners on a third type of cell (lung cancer cells) that they hadn't used for training, the results were mixed. Some winners worked, some didn't. This shows that the "perfect key" depends heavily on the specific "lock" (cell type).

Why This Matters

This paper is a game-changer because it flips the script.

• Old Way: "I think this shape will work. Let me build it and test it." (Slow, limited).
• New Way: "Here are a million random shapes. Let the cells tell me which ones work." (Fast, explores the whole universe of possibilities).

The Big Picture:
Imagine you are trying to find the best shoe for a marathon. Instead of designing one shoe and hoping it fits, you throw 10,000 different shoes into a room with 10,000 runners. You see which shoes the runners actually wear and run in. Then, you make more of those specific shoes and repeat the process. Eventually, you end up with the perfect marathon shoe without ever having to design it yourself.

This method could help scientists quickly discover the perfect DNA carriers to deliver cancer drugs directly into tumor cells, bypassing the need for trial-and-error design. It turns the discovery process from a slow design project into a fast, automated evolutionary race.

Technical Summary

1. Problem Statement

DNA nanotechnology offers precise, biocompatible structures for targeted drug delivery. However, current discovery methods rely on the rational design and testing of individual nanostructures one by one. This approach is inefficient for exploring the vast design space of DNA nanostructures, where size, shape, and mechanical properties can be tuned independently. Furthermore, traditional DNA origami and tile-based assemblies are composed of multiple strands that disassemble upon heating, making them incompatible with the amplification and sequencing required for evolutionary selection strategies (similar to SELEX for aptamers). There is a critical need for a scalable, high-throughput method to screen large libraries of DNA nanostructures to identify those with preferential cellular uptake for specific cell types.

2. Methodology

The authors developed a novel evolutionary selection pipeline that combines a unique library design with iterative cellular screening and high-throughput sequencing.

• Library Design ("Structure Genome"):

  • Instead of multi-strand origami, the library consists of DNA nanostructures that fold from a single long single-stranded DNA (ssDNA) molecule.
  • The library was generated by ligating a diverse pool of 36 pre-assembled DNA fragments (ranging from linear duplexes to junctions, hairpins, and kissing loops).
  • Each fragment was designed with specific sticky-end overhangs (AGT/ACT) to ensure directional ligation.
  • Crucially, the library includes an "amplification loop" containing primer binding sites and two 10-nucleotide Unique Molecular Identifiers (UMIs) at the ends. This allows every unique nanostructure to be tracked as a "genome" via PCR and sequencing.

• Library Preparation & Purification:

  • Fragments were ligated using T4 DNA ligase.
  • Exonuclease III (Exo III) treatment was applied to digest linear or partially ligated fragments (which have free 3' ends), enriching the pool for covalently closed, topologically complete nanostructures.
  • The library was PCR amplified, purified, and converted to ssDNA using Lambda exonuclease (which digests the 5'-phosphorylated strand) to allow refolding into native conformations for cell incubation.

• Iterative Cellular Selection:

  • Cell Lines: HEK293T (human embryonic kidney) and RAW264.7 (mouse macrophage-like) cells were used.
  • Process: The ssDNA library was incubated with cells for 4 hours. Cells were washed to remove surface-bound DNA, lysed, and internalized DNA was recovered.
  • Amplification Cycle: Recovered DNA was PCR amplified, converted back to ssDNA, refolded, and re-incubated with cells. This cycle was repeated for 10 rounds.
  • Controls: An "empty selection" control (10 rounds of amplification without cells) was performed to rule out PCR bias.

• Sequencing & Analysis:

  • Libraries from initial and selected rounds were sequenced using both Oxford Nanopore Technologies (ONT) (for long reads) and Illumina (for high accuracy).
  • Bioinformatics tools (cONcat) were used to map reads back to the initial fragment library to reconstruct the "structure genome" composition.
  • UMIs were used to count unique structure abundance and track enrichment.
  • Reads aligning to the human reference genome (GRCh38) were filtered out to remove genomic contamination.

• Validation:

  • Top candidates identified from sequencing were synthesized as individual constructs.
  • Cellular uptake was quantified using Flow Cytometry (Cy5 fluorescence) and Confocal Microscopy in HEK293T, RAW264.7, and a non-selected cell line (A549).

3. Key Contributions

• Novel Library Architecture: Introduced a "structure genome" approach where diverse nanostructures are encoded in a single strand, enabling them to survive PCR amplification and sequencing while retaining structural information.
• Selection-Based Discovery: Successfully inverted the traditional design process by using cellular uptake as a selection pressure to evolve DNA nanostructures from a random library.
• Cell-Type Specificity: Demonstrated that the selection process yields different structural preferences depending on the cell type (e.g., HEK293T vs. RAW264.7).
• Scalable Pipeline: Established a generalizable workflow for engineering DNA nanocarriers tailored to specific biological contexts without prior rational design.

4. Key Results

• Enrichment Dynamics:
  • After 10 rounds of selection, specific structural motifs were significantly enriched.
  • Fragment Preferences: Both cell lines showed a strong preference for linear fragments and negative selection against kissing loop segments.
  • Cell Differences: HEK293T cells showed a more stringent selection process, resulting in a distinct shift toward shorter, specific structures (average size ~200–500 bp). In contrast, RAW264.7 (macrophage) cells exhibited "promiscuous" uptake, showing less dramatic shifts in length distribution, suggesting they internalize a broader range of structures.
• Candidate Identification:
  • Several candidates (e.g., H10A, H10B from HEK293T; R10B from RAW264.7) were identified with high UMI counts and specific enrichment profiles.
  • Some highly enriched sequences were found to contain human genomic fragments (chimeras), highlighting the complexity of the selection environment, though synthetic candidates were also successfully isolated.
• Validation of Uptake:
  • Flow Cytometry: Selected candidates showed significantly higher internalization (Cy5+ cells) compared to random controls in their respective selection cell lines. For example, H10A and H10B showed ~41% uptake in HEK293T, while R10B showed high uptake in RAW264.7.
  • Specificity: Candidates selected in HEK293T generally performed better in HEK293T than in RAW264.7, and vice versa, confirming cell-type specificity.
  • Non-Selected Cell Line (A549): Uptake was generally low in A549 cells. Interestingly, a "Crude Control" candidate (which was not enriched in the selection) showed the highest uptake in A549, suggesting that the selection process might miss some effective structures due to stability or amplification biases.
• Subcellular Localization:
  • Confocal microscopy revealed heterogeneous uptake in HEK293T (some nuclear localization, some cytosolic).
  • RAW264.7 cells showed uniform, cytosolic uptake with no nuclear entry, consistent with their phagocytic nature.

5. Significance

This work represents a paradigm shift in DNA nanotechnology from rational design to evolutionary selection.

• High-Throughput Exploration: It enables the screening of thousands of structural variants simultaneously, overcoming the bottleneck of testing designs one by one.
• Tailored Nanomedicine: The ability to select for cell-type-specific uptake is crucial for developing targeted drug delivery systems that minimize off-target effects.
• Future Applications: The framework is adaptable. Future iterations could integrate functional groups (e.g., aptamers, peptides) directly into the library or perform selection in vivo to identify structures capable of targeting specific tissues or tumors.
• Overcoming Limitations: By using single-strand folding, the method bypasses the thermal instability issues of traditional origami, making the structures compatible with standard molecular biology tools (PCR, sequencing).

In conclusion, the authors have successfully demonstrated a scalable pipeline to discover DNA nanostructures with optimized cellular internalization properties, paving the way for the systematic development of programmable nanomedicines.