Expanding the Drug Delivery Toolbox

KdT Ventures
9 min readSep 15, 2022


Today our ability to treat many diseases is limited only by our ability to effectively deliver genetic medicines.

Patients treated with genetic medicines receive nucleic acids, either DNA or RNA, to fix, replace or fine tune the expression of certain genes. By supplying patient cells with new functional copies of genes or fixing mutated genes, genetic medicines provide the biological “source code” to not only treat, but to cure disease. These programmable medicines have the potential to address unmet needs for millions of patients, including those with rare genetic diseases for which few effective treatments are available.

With 8 FDA approved therapies and 120 more in at least phase II trials, viral vectors are currently the standard delivery system for genetic medicines, of which, the adeno-associated virus (AAV) is the most common. Viral-delivered gene therapies have demonstrated clinical benefit across a range of genetic diseases; however, their ability to unlock the full potential of genetic medicine faces significant limitations:

1. Immunogenicity: Viral delivery systems encapsulate therapeutic cargos (proteins, DNA, and/or RNA) inside a viral capsid which the immune system can recognize as foreign. Immune recognition and elimination of viral capsids limits both initial delivery of the drug and the ability to re-dose it.

2. Safety: Over a third of AAV gene therapy clinical trials have reported serious adverse events (SAEs) including liver toxicity and severe inflammation. These SAEs were caused by the non-native nucleic acids, the viral capsid, or the immune system’s response to these components. Tragically, just this month, two children died of acute liver failure within 5 weeks of taking Novartis’s AAV-gene therapy, Zolgensma, for spinal muscular atrophy.

3. Size/Delivery Capacity: Delivery by viral vectors presents a packing problem: genetic medicine cargos must fit within the viral capsid. 83% of genes are not viable for AAV gene therapy because they exceed the AAV’s ~4.2 kb cargo capacity.

4. Cell Specific Delivery: Delivery by viral vectors is still too limited to certain privileged cell types (for example, the eye and the liver). Our collective ability to deliver therapeutic payloads to other cell types (and only those specific cell types) remains more aspirational than resolved.

While viral vectors have effectively been used to deliver nucleic acid-based drugs to patients, they are not one size fits all. Next-generation biologics and genetic medicines need a robust toolbox tailored to address specific diseases, once the “low hanging fruit” are all gone. They’ll need a delivery platform that is customizable for targeted cell delivery, has a favorable safety profile, and is adaptable to a wide breadth of therapeutic modalities.

Non-viral vectors have the potential to provide benefits where viral vectors are limited. The genes driving Parkinson’s Disease, Duchenne Muscular Dystrophy and Stargardt Disease all exceed the capacity of a single AAV vector, but not non-viral vectors. Further, non-viral systems can deliver a greater level of therapeutic diversity (e.g., DNA, RNA, entire proteins, and protein-nucleic acid complexes) than even the best current viral vectors. The cargo capacity and flexibility of non-viral vectors opens the door to pursue “undruggable” targets and impact previously untreatable diseases. For example, autoimmune diseases or cancers are prime targets for non-viral vectors given their need for redosing and more tightly controlled therapeutic indexes. Non-viral delivery systems are a potential swiss army knife offering the flexibility needed for next-gen therapies.

The most promising non-viral delivery systems currently fit into three main classes:

References: 1 & 2. Figure 1. Breakdown of non-viral delivery classes

While we have identified and celebrated the virtues of these non-viral delivery systems, there are still significant challenges to overcome prior to realizing their vast potential:

  1. Biodistribution & Targeted Cell Delivery: Drugs must be delivered to not only the correct tissues but the correct cells within those tissues to be effective. The challenges of drug delivery are similar to those of parcel delivery: biodistribution is being in the right zip code while targeted cell delivery is having the right key to enter a specific house. Controlled biodistribution is a prerequisite for targeted cell delivery. Overcoming the challenges of targeted cell delivery — matching molecules on the surface of the delivery vector with receptors on the targeted cell — is relatively easy in theory. Once we identify a molecule-receptor pair of interest, targeting molecules (e.g., scFvs, VHHs, or DARPins) can be engineered and loaded onto the surface of the delivery vector. The real challenge is how to meaningfully shift vector biodistribution — where in the body the vectors accumulate. Some delivery vectors accumulate in the liver, contributing to liver toxicity and preventing delivery of the therapeutic cargo to the appropriate tissues or alternatively, delivery to the wrong tissues by proximity alone. Ultimately, if the therapeutic cargo can’t be delivered to the specific cells where it is needed, then the drug becomes worthless– or, worse yet, dangerous.
  2. Efficient Cargo Release: Hitting the desired cells alone doesn’t guarantee a therapeutic effect. The cargo must be released into the correct compartment of the target cell to perform its therapeutic function. Cargoes taken up by target cells can be trafficked through and stored in cellular sub-compartments. An Alnylam study estimated that only 1–2% of siRNA cargo molecules delivered by an LNP are ultimately released into the cytosol. Uptake of therapeutic cargoes into sub-compartments that either sequester them from appropriate targets or lead to their premature degradation can negatively impact the therapeutic effects of cargoes, even when they are delivered to the targeted cell. Controlling cargo release, trafficking, and cellular persistence will be critical for maximizing efficacy and dosing.
  3. Immunogenicity & Toxicity: The structural components of EVs, derived from human cells, and LNPs, synthetically produced but inspired by human liposomes, are expected to be less immunogenic than those of VLPs and viral vectors. However, the safety profiles of EVs and LNPs are not yet fully characterized. While the COVID mRNA vaccines have an excellent track record of efficacy and safety, greater than expected local inflammation and flu-like symptoms have been observed in humans compared to animal models following LNP-mRNA vaccination. Immune reactions may limit the ability to redose drugs delivered using these vectors– effectively making their immediate adaptation to treat chronic disease impossible. A recent study determined that both the mRNA and the lipid components of LNP mRNA vaccines can contribute to this immune response. Innovative solutions to minimize potential safety risks while maintaining targeting and cargo delivery will be needed as non-viral vectors are developed into drug delivery systems.
  4. Complex Manufacturing Processes: The manufacturability of any therapy can be a limiting factor for its widespread application. If the cost of goods and services are too high, adoption will be too low. Meanwhile, inconsistent production and purification compromises the ability to reliably deliver therapeutic doses and threatens regulatory approval. LNP formulations require complex chemical reactions and physical manipulation to 1) form consistent nanostructures, 2) add on functional moieties, and 3) load therapeutic cargo. In contrast, EVs offer a genetically controlled approach to manufacture delivery vehicles by using producer cell lines in which EV characteristics such as lipid composition and surface protein content can be genetically programmed. Cargo loading into EVs can also be programmed, or at least influenced, through overexpression, epitope tagging, or bifunctional molecule tethering. However, isolation and characterization of the produced EV populations remains a challenge. VLP manufacturing relies on genetically programmed phage cell lines, but is the least well developed and has the least biotech activity compared to EVs and LNPs. Manufacturing will continue to limit the efficacy of non-viral delivery until our ability to modulate, characterize, and purify these vehicles improves.

Among the three classes of non-viral delivery systems, EVs are uniquely “platformable”. Because they originate from genetically programmed producer cell lines, EVs are theoretically easier to engineer with specific characteristics than LNPs, which must be synthetically formulated using rigid chemistry. Genetically programmed EV biogenesis lends itself to a model of iterative, AI-assisted design-test-learn cycles to efficiently optimize EVs for delivery.

Of the three non-viral delivery modalities discussed above, we believe EVs have the greatest potential to fill the void left by viral vectors. However, in addition to the obstacles described above, there other potential hurdles unique to EVs that will require further optimization:

  1. Shifting Biodistribution Profile: To deliver EVs to specific cells, many companies express targeting molecules (e.g., proteins, viral fusogens) on EV surfaces. For example, Codiak has identified novel classes of exosome-associated proteins which they use as scaffolds to display various targeting ligands on EV surfaces. But targeting molecules is only part of the battle. To direct EV biodistribution Codiak is utilizing a “compartmental dosing approach” to administer EV therapies through a variety of different routes (e.g., intravenous, intracranial, intramuscular etc…). By selectively matching targeting molecules with a specific route of administration, Codiak aims to locally drive biodistribution to specifically target diseased tissues. While, their myeloid-targeted, oncology-focused pipeline and data to date don’t suggest Codiak has avoided the macrophage clearance issue that has plagued EVs, it might not have to for Codiak to be successful in the clinic. Further work to characterize and systematically program EV features that influence their biodistribution and bioavailability will be critical in unleashing their potential as a delivery vector.
  2. Refining CMC Standards and Processes: EVs are extremely complicated delivery vectors. They consist not only of the drug payload, but also the macromolecular shell which carries them. How do we more efficiently build and load these carriers in a repeatable, scalable way? Syenex, a startup out of the Leonard Lab at Northwestern, has developed a proprietary approach to load tagged therapeutic cargoes and surface targeting molecules simultaneously via small molecule-mediate dimerization. Dimerization of the two proteins occurs when each half of the dimer (the cargo and the surface molecule) binds to one side of a bifunctional small molecule. This small molecule-regulated loading is readily applicable to large scale biomanufacturing processes. Evox Therapeutics is a more established EV company that applies models from conventional biologics manufacturing to optimize exosome manufacturing lines, but they are keeping the specifics of their proprietary approach close to the vest. Ultimately, manufacturing and purification will drive the transition of EVs from biological phenomenon to a reproducible delivery system.
  3. Clinical Validation: Finally, EVs need to demonstrate clinical success. Codiak is the most advanced EV company identified here, but even Codiak’s most advanced programs haven’t progressed beyond phase 1 clinical trials. In patients, EVs must live up to their potential to selectively target diseased cells, decrease immunogenicity compared to viral vectors, and enable re-administration of genetic medicines while maintaining an impeccable safety profile.

Every month new biotech companies emerge from stealth aiming to solve non-viral delivery. Propelled by an increasingly interested set of capital investors, the pace of innovation has the potential to be exponential. Below is just a sample of groups focused on solving delivery with non-viral vectors.

Figure 2. Biotech companies innovating in the non-viral delivery space.

Non-viral delivery systems currently teeter at a critical inflection point. We’ve come a long way in understanding what components are required to build a delivery system, but that knowledge has yet to translate into safe, effective non-viral delivered therapies. Winners in the non-viral delivery space will build rational solutions to deliver the next generation of drugs and biologics. They will combine advances in DNA synthesis, high-throughput screening and ML/AI to query the entire search space of possible vector designs, akin to Dyno Therapeutics’ CaspidMap Platform, which led the wave in AAV vector-fitness mapping and design. With an efficient design-test-learn cycle, they will be able to understand how each individual piece of the non-viral vector contributes to its function in vivo. The winners will be able to iterate this cycle in the fastest, most efficient manner to optimize for biodistribution, targeted cell delivery, efficient cargo loading and release, and safety.

While today’s viral delivery methods are equipped to treat a subset of disease types, the future biotech delivery toolbox needs a more robust toolset. We believe that non-viral vectors, especially EVs, are primed to be the swiss army knife of delivery and usher in this next generation of therapies.

Authors: Emilio Ferrara and Tamara Vital, KdT Ventures



KdT Ventures

KdT is the standard for early-stage science venture investing. We help founders and their companies re-architect the world at a molecular level.