Structural dynamics insights into principles underlying the fitness of new broadly potent AAVs

By integrating a single-cell engineering workflow with molecular dynamics simulations, this study elucidates the structural dynamics mechanism—specifically enhanced receptor affinity and regulated heparan sulfate binding—that underpins the broad potency of the engineered AAV vector ATX002, thereby establishing a new framework for rational gene therapy vector design.

The Gist

The Big Picture: Fixing the "Delivery Truck" Problem

Imagine you want to deliver a life-saving package (a gene therapy) to a specific house in a city (a cell in your body). The current delivery trucks (viruses used in gene therapy) are okay, but they have two big problems:

1. They are slow and inefficient: You need to send thousands of trucks just to get one package to the right door.
2. They cause traffic jams: Because you have to send so many trucks, they clog up the streets, causing accidents and angering the neighbors (the immune system), which can be dangerous.

Scientists wanted to build a super-truck that is fast, efficient, and can deliver packages to the right house without causing a traffic jam. This paper is about how they built that super-truck, named ATX002, and figured out exactly why it works so well.

---

Part 1: The "Talent Show" for Viruses

To find the best truck, the scientists didn't just guess. They held a massive Talent Show (called scAAVengr-HUnT).

• The Contestants: They created a library of thousands of slightly different virus trucks. Think of it like changing the bumper stickers, the paint job, or the shape of the headlights on a standard truck.
• The Stage: Instead of testing them in a lab dish, they injected these trucks directly into the eyes of monkeys (who have eyes very similar to humans).
• The Judges: They used a high-tech camera system (single-cell RNA sequencing) to look at every single cell in the retina. They asked: "Which truck got inside? Which one delivered its package? Which one reached the back of the room?"

The Winner: One truck, ATX002, was the clear champion. It didn't just win; it dominated. It delivered its cargo 14 times better than the current best truck used in hospitals. It reached the front of the eye (the fovea) and the back (the periphery) equally well, something other trucks struggle to do.

Part 2: The "Magic Loop" on the Truck

Once they found the winner, they wanted to know: What makes ATX002 so special?

They took a super-microscope (Cryo-EM) to look at the truck's structure. They found that the main body of the truck was identical to the old models. The only difference was a tiny, flexible loop of amino acids (a little piece of string) sticking out of the top.

• The Analogy: Imagine the truck has a standard door handle. ATX002 has a special, custom-made handle attached to it.
• The Discovery: This little loop wasn't just decoration; it was the secret weapon.

Part 3: The "Double-Agent" Secret (How it Works)

This is the most fascinating part. The scientists used a computer simulation (Molecular Dynamics) to watch the truck move in slow motion, like a high-speed video game. They discovered that this little loop acts like a chameleon or a double-agent. It has two different jobs depending on where the truck is:

Job 1: The "Stealth Mode" (In the open)
When the truck is floating in the eye fluid, it needs to avoid getting stuck on walls or sticky barriers (like the Inner Limiting Membrane).

• The Mechanism: The loop folds in on itself. The positive and negative charges inside the loop cancel each other out, like magnets snapping together.
• The Result: The truck becomes "invisible" to sticky barriers. It glides right past them without getting stuck, allowing it to travel deep into the eye.

Job 2: The "Key Mode" (At the door)
When the truck finally arrives at the target cell door, it needs to unlock it.

• The Mechanism: The loop unfolds! The charges that were hiding now pop out and grab onto the cell's receptor (the door handle) with a strong, magnetic grip.
• The Result: The truck latches on tightly and opens the door to deliver the medicine.

The "Bifunctional" Magic:
Most trucks are either good at gliding (but bad at grabbing) or good at grabbing (but get stuck on walls). ATX002 is a bifunctional truck: it knows when to be slippery and when to be sticky. It's like a spy who wears a disguise to sneak past guards, but then reveals their ID badge to get into the VIP room.

Part 4: Does it work on Humans?

The scientists tested this super-truck in three different "cities":

1. Monkeys: It worked perfectly, just like in the talent show.
2. Mice: It worked even better than the old trucks.
3. Human Cells: They grew human retinal cells in a dish (like a mini-eye in a petri dish). ATX002 successfully delivered the medicine to these human cells, proving it's ready for the real world.

They also tested it in the brain. Usually, getting medicine into the brain is incredibly hard because of a protective shield. ATX002 managed to cross this shield and deliver medicine to brain cells, outperforming the current gold standard (AAV9).

The Takeaway

This paper is a breakthrough because it doesn't just say, "We found a better truck." It explains the physics of why it works.

• Old way: "Let's try random changes and hope one works."
• New way: "Let's design a truck with a loop that knows how to hide its stickiness until it reaches the door."

This discovery gives scientists a new blueprint. Instead of just guessing, they can now design future gene therapies with specific "smart loops" that make them safer, more effective, and capable of treating diseases in the eye and brain that were previously impossible to reach.

Technical Summary

1. Problem Statement

Adeno-associated viruses (AAVs) are the leading platform for gene therapy, yet current clinical-stage vectors suffer from significant limitations:

• Inefficiency: They require high doses to achieve therapeutic transduction, which is associated with immune responses, toxicity, and severe adverse events (including fatalities in clinical trials).
• Delivery Barriers: In ocular gene therapy, delivering vectors to the outer retina (photoreceptors) via intravitreal (IVT) injection is challenging due to structural barriers like the inner limiting membrane (ILM). Subretinal injections are effective but invasive and carry risks of iatrogenic damage.
• Lack of Mechanistic Insight: While engineering has produced improved vectors (e.g., 7m8), the biophysical mechanisms underlying their enhanced "fitness" (transduction efficiency and tropism) remain poorly understood. This lack of mechanistic guidance limits the rational design of next-generation vectors.

2. Methodology

The authors employed a multi-disciplinary approach combining high-throughput biological screening, structural biology, and computational biophysics:

• scAAVengr-HUnT Platform: A novel workflow for Single-cell AAV Engineering with High Unbiased Throughput.
  • Library Construction: Created diverse AAV2-based capsid libraries with a 6-amino acid insertion (TC6) at position 588 of the VP1 protein (an exposed loop interacting with heparan sulfate).
  • In Vivo Selection: Libraries were injected intravitreally into Non-Human Primate (NHP) eyes.
  • Single-Cell Resolution: Retinas were dissociated, and single-cell RNA sequencing (scRNA-seq) was performed to quantify mRNA expression of the transgene (eGFP) and barcode representation for each variant. This allowed for head-to-head comparison of thousands of variants across specific retinal cell types and regions (fovea vs. periphery).
• Structural Biology:
  • Cryo-EM: Determined the atomic structure of the top-performing variant (ATX002) at 1.9 Å resolution.
  • Comparison: Compared ATX002 structures against parental AAV2 and the clinical standard 7m8.
• Computational Biophysics (Molecular Dynamics - MD):
  • Performed 1-µs all-atom MD simulations on 3-fold symmetric capsid segments (3 VP3 proteins).
  • Simulated variants in two states: Isolated (to study intraloop interactions and heparan sulfate binding potential) and Complexed with AAVR (the universal AAV receptor) to study receptor engagement.
  • Analyzed hydrogen bonding, salt bridges, and contact frequencies between the capsid loops and receptors.

3. Key Contributions

• Discovery of ATX002: Identification of a new "best-in-class" AAV vector (ATX002) containing a specific 9-residue insertion (LAEHQRTPA) that outperforms current clinical standards.
• Cross-Species and Cross-Tissue Validation: Demonstrated that ATX002's potency translates across species (NHP, mouse, human organoids) and tissues (retina and CNS).
• Mechanistic Breakthrough: Provided the first mechanistic explanation for engineered AAV fitness using structural dynamics, revealing a bifunctional molecular mechanism.
• Integration of Workflows: Established a pipeline integrating sequence-level analysis, single-cell in vivo screening, and computational biophysics to guide AAV engineering.

4. Key Results

A. Performance of ATX002

• Potency: ATX002 is 14-fold (1.16 log10) more potent than the clinical vector 7m8 in NHP retinas. It transduces significantly more cells across the entire retina (pan-retinal), including the fovea and periphery.
• Therapeutic Efficacy: When delivering the therapeutic transgene RS1 (for X-linked retinoschisis), ATX002 showed up to 328-fold higher expression compared to AAV8 (a vector currently in clinical trials).
• Cross-Species Translation:
  • Mouse: Outperformed AAV2 and 7m8 in IVT injections.
  • Human Organoids: Efficiently transduced photoreceptors in iPSC-derived human retinal organoids and iRPE cells.
• CNS Delivery: Following intracisterna magna (ICM) injection, ATX002 demonstrated superior transduction of the CNS (cerebellum, thalamus, brainstem, hippocampus) in both mice and NHPs compared to AAV9 (the current gold standard for CNS delivery). It showed high neuronal specificity.

B. Structural Insights

• Cryo-EM: The capsid structure of ATX002 is essentially identical to AAV2 and 7m8 (RMSD ~0.77 Å). The 9-residue insertion forms a flexible loop that does not alter the global capsid fold.
• MD Simulations & The Bifunctional Mechanism:
  • Receptor Binding (AAVR): High-fitness variants (ATX002, 7m8) utilize inserted charged residues (arginine and glutamate) to form extensive salt bridges and hydrogen bonds with the AAV receptor. This enhances binding affinity compared to AAV2.
  • Heparan Sulfate (HS) Regulation: A critical finding is that while the inserted arginine could theoretically increase non-productive binding to HS (a barrier in the eye), the inserted glutamate forms a salt bridge with the arginine within the loop.
  • The Mechanism: This intraloop charge pairing sequesters the positive charge, shielding it from HS and preventing unproductive retention in the ILM. However, upon encountering the AAV receptor, the loop repositions, allowing the charged residues to engage the receptor effectively.
  • Failure Modes: Variants with shifted insertions (e.g., ATX002_ins_589) disrupt this shielding or receptor contact, leading to poor fitness. Variants with nonpolar inserts (var16) fail to enhance receptor binding.

5. Significance

• Clinical Impact: ATX002 offers a solution for treating retinal diseases requiring widespread delivery (e.g., retinitis pigmentosa, macular degeneration) at significantly lower doses, potentially mitigating the toxicity and immune responses associated with high-dose AAV therapies.
• Paradigm Shift in Engineering: This work moves AAV engineering beyond simple sequence optimization. It demonstrates that structural dynamics (how the capsid moves and interacts with receptors/barriers) are critical determinants of fitness.
• Future Design Principles: The study establishes a general design rule: successful engineered loops must balance enhanced receptor affinity with shielding of non-productive interactions. This "bifunctional" principle provides a blueprint for designing safer, more efficient vectors for gene therapy across various tissues and species.
• Methodological Advance: The scAAVengr-HUnT platform combined with MD simulations offers a powerful, scalable framework for the rational design of viral vectors, bridging the gap between high-throughput screening and atomic-level mechanistic understanding.