- The Delivery Problem: Why It’s Hard
- What Form to Deliver: DNA, mRNA, or RNP
- Viral Vectors: AAV
- Lipid Nanoparticles (LNPs)
- Electroporation
- Other Methods: Microinjection, Nanoparticles
- Ex Vivo vs In Vivo: Two Fundamentally Different Strategies
- Delivery for Specific Tissues: Liver, Eye, Brain, Lung
- How Casgevy Solved Delivery for Sickle Cell
- The Story: The Surgical Team and the Locked Building
Section 1 — The Delivery Problem: Why Getting CRISPR Into Cells Is Hard
Think about what you are trying to do. You have a large protein (Cas9, ~160 kilodaltons) and an RNA molecule (the guide RNA), and you need to get both of them inside a specific cell type, through several layers of biological barriers, in sufficient quantity to edit the genome, without killing the cell, without triggering an immune response, and — if you are treating a patient — without damaging unintended tissues.
The cell membrane is the first barrier. It is a lipid bilayer — a two-layer sheet of fat molecules — that is specifically designed to keep large molecules out. Proteins, RNA, and DNA are all hydrophilic (water-loving) molecules that cannot spontaneously cross the hydrophobic (fat) interior of the membrane. Without a delivery vehicle, CRISPR components injected into a bloodstream will degrade within minutes, never enter a cell, and produce no editing whatsoever.
Beyond the membrane, there are additional barriers: endosomes (cellular compartments that trap ingested molecules and can degrade them before they reach the nucleus), the nuclear envelope (a second membrane around the nucleus that the Cas9-gRNA complex must penetrate), and the nuclease activity of the cytoplasm (enzymes that chew up naked RNA and DNA). Delivery is not a single barrier to cross — it is a gauntlet.
Section 2 — What Form to Deliver: DNA, mRNA, or Ribonucleoprotein
Before choosing a delivery vehicle, you must choose what molecular form to deliver the CRISPR system in. The same Cas9 + guide RNA combination can be packaged and delivered as three fundamentally different molecular species, each with distinct advantages, kinetics, and applications.
Both Cas9 and guide RNA are encoded as DNA. The cell transcribes Cas9 mRNA, translates Cas9 protein, and transcribes the guide RNA from the DNA template. Editing is slow to start (hours to days for protein to accumulate) but sustained (the DNA template persists in the cell). Plasmid DNA can integrate into the genome, maintaining Cas9 expression long-term — a safety risk for therapeutic use.
Cas9 is delivered as synthetic mRNA (the same technology as COVID-19 vaccines); the guide RNA is delivered separately as synthetic RNA. The cell translates the mRNA into Cas9 protein rapidly (within hours), edits the genome, and then the mRNA degrades (within days). No DNA integration risk. This is the approach used in Intellia’s NTLA-2001 therapy delivered via lipid nanoparticles to the liver.
Pre-assembled Cas9 protein already complexed with the guide RNA — ready to cut immediately upon entering the nucleus. No transcription or translation required. Fastest editing kinetics. Shortest duration of Cas9 activity (protein degrades within hours to days). Lowest off-target activity because Cas9 is present transiently. This is the gold standard for ex vivo editing (electroporation of cells removed from the body). Used in Casgevy (sickle cell therapy).
Section 3 — Adeno-Associated Virus (AAV): The In Vivo Gold Standard
Adeno-associated viruses are small, naturally occurring viruses that infect humans but cause no known disease. They are the most widely used viral vector for in vivo gene therapy — delivering genetic material directly into tissues in a living organism — for several compelling reasons: they are non-integrating (they exist as episomal circles in the nucleus rather than inserting into the genome), they have low immunogenicity compared to other viruses, they can transduce non-dividing cells, and their tropism (which tissues they preferentially infect) can be controlled by choosing different serotypes.
An AAV vector is a gutted version of the natural virus: all viral genes are removed and replaced with your cargo (the CRISPR components), leaving only the inverted terminal repeats (ITRs) that the virus uses for replication and packaging. The result is a non-replicating particle that can infect cells and deliver DNA but cannot reproduce, spread, or cause disease.
AAV Serotypes: Choosing the Right Tissue Target
There are over 12 naturally occurring AAV serotypes (AAV1 through AAV12, plus dozens of engineered variants), each with different cell surface receptor preferences and therefore different tissue tropisms. Selecting the right serotype is as important as selecting the right guide RNA — the wrong serotype will simply fail to infect your target tissue.
The Size Problem: AAV and CRISPR Don't Fit Well Together
AAV has a packaging limit of approximately 4.7 kilobases (kb) of DNA. SpCas9 alone requires ~4.2 kb to encode. Adding the guide RNA expression cassette and the AAV ITRs pushes the total to ~5.5 kb — exceeding the packaging limit. This is a fundamental physical constraint that the field has addressed in several ways:
SaCas9 (~3.2 kb coding sequence) fits comfortably in a single AAV with the guide RNA. CjCas9 (~2.9 kb) fits even more easily. Trade-off: different PAM requirements and potentially lower efficiency than SpCas9.
SpCas9 is split into N- and C-terminal halves, each packaged in a separate AAV vector. Both vectors infect the same cell; split-intein sequences splice the halves together into functional Cas9. Efficient in vivo but requires co-delivery of two vectors.
Engineered Cas9 variants with internal deletions that reduce size while maintaining activity. The REC2 domain of SpCas9 has been partially deleted without abolishing function, bringing the coding sequence closer to 3.5 kb.
For liver targeting — the most common therapeutic application — LNPs delivering Cas9 mRNA + gRNA can achieve equivalent or better editing without any size constraint. Intellia and others have abandoned AAV for liver targets in favour of LNPs.
Section 4 — Lipid Nanoparticles: The mRNA Vaccine Technology Applied to CRISPR
Lipid nanoparticles (LNPs) are the delivery technology that brought COVID-19 mRNA vaccines to billions of people. The same technology — with modifications — is now being applied to deliver CRISPR components. LNPs are tiny fat bubbles, typically 80–150 nanometres in diameter, that encapsulate their cargo (mRNA, gRNA, or both) and protect it from degradation until it reaches the target cell.
The key innovation in therapeutic LNPs is the ionisable lipid component. Unlike older cationic (permanently positively charged) lipids, ionisable lipids are neutral at physiological pH but become positively charged in the acidic environment of endosomes after cellular uptake. This charge change disrupts the endosomal membrane, releasing the LNP contents into the cytoplasm — solving the endosomal escape problem that plagued earlier lipid delivery systems.
LNP Composition: The Four-Component System
The workhorse component. Neutral at pH 7.4 (blood), charged at pH 5.5 (endosome). Drives endosomal escape. Examples: DLin-MC3-DMA (Onpattro), ALC-0315 (Pfizer vaccine), SM-102 (Moderna vaccine). The specific structure of the ionisable lipid strongly determines LNP efficacy.
A phospholipid that helps form the bilayer structure of the LNP and promotes endosomal escape by destabilising the endosomal membrane. Also contributes to cell membrane fusion.
Provides structural stability to the LNP and promotes membrane fusion. Enhances cellular uptake and endosomal release. Cholesterol is a natural component of cell membranes, which reduces toxicity.
Polyethylene glycol conjugated to a lipid forms a protective hydrophilic shell around the LNP. Prevents aggregation, reduces protein adsorption (opsonisation), and extends circulation half-life in the blood. Too much PEG reduces cellular uptake; too little reduces stability.
The Liver Problem: LNPs Go Where ApoE Takes Them
After intravenous injection, LNPs circulate in the blood and accumulate primarily in the liver. This happens because LNPs adsorb apolipoprotein E (ApoE) from the blood, and hepatocytes (liver cells) express high levels of the LDL receptor that recognises ApoE-decorated particles. The result is highly efficient liver transfection — which is excellent for liver diseases (transthyretin amyloidosis, haemophilia, hypercholesterolaemia) but limits LNP utility for other tissues.
Achieving tissue-specific delivery beyond the liver with LNPs is an active and intensely competitive research area. Strategies include: incorporating targeting ligands (antibodies, peptides) on the LNP surface, altering lipid composition to change organ tropism, and developing selective organ targeting (SORT) LNPs by adding a fifth charged lipid component. Early data suggests that SORT-LNPs can redirect editing to the spleen, lungs, and other organs — but none are yet in clinical use for CRISPR.
Section 5 — Electroporation: The Gold Standard for Ex Vivo Editing
Electroporation uses brief, controlled electrical pulses to temporarily create nanoscale pores in the cell membrane, allowing large molecules — including Cas9 RNP complexes, mRNA, or plasmid DNA — to enter the cell directly. It is not elegant, but it is extraordinarily effective. Electroporation is the delivery method of choice for ex vivo cell editing, and it is the method used in the approved sickle cell therapy Casgevy.
Modern electroporation systems (Lonza Nucleofector, MaxCyte STX, Thermo Fisher Neon) are highly optimised for specific cell types. They use precisely calibrated voltage, pulse duration, and pulse number to maximise delivery efficiency while minimising cell death. For haematopoietic stem cells (HSCs) — the target of Casgevy — optimised electroporation conditions deliver Cas9 RNP to more than 90% of cells with greater than 70% cell viability.
| Parameter | Effect on Outcome | Typical Range |
|---|---|---|
| Voltage | Higher voltage creates more/larger pores but increases cell death | 100–2,000 V depending on cell type |
| Pulse duration | Longer pulses allow more cargo entry but increase toxicity | 0.1–30 ms |
| Buffer composition | Cell-type specific buffers dramatically affect viability and efficiency | Varies by manufacturer and cell type |
| RNP concentration | More RNP = higher editing but potentially more off-target and toxicity | 1–10 μM Cas9, 2–3 × molar excess gRNA |
| Cell density | Crowding affects electrical field distribution and recovery | 106–108 cells/mL depending on format |
Section 6 — Other Delivery Methods: Microinjection, Nanoparticles, and More
Microinjection
Microinjection uses a glass needle thinner than a micrometre to physically inject CRISPR components directly into a cell’s nucleus or cytoplasm. It is the most precise and most efficient delivery method — essentially 100% of injected cells receive the cargo. It is also the method used for editing mouse embryos, fish embryos (zebrafish), and human embryos in research settings.
The obvious limitation: microinjection requires a skilled operator and a micromanipulator apparatus, and can only process one cell at a time. It is completely impractical for the millions of cells needed for therapeutic applications, but remains the gold standard for single-cell research editing and generating animal models.
Polymer and Inorganic Nanoparticles
Polymer nanoparticles (polyethylenimine, PLGA, chitosan), gold nanoparticles conjugated to CRISPR components, and silica nanoparticles are all being developed as non-viral delivery alternatives. These approaches offer the possibility of different tissue targeting profiles than LNPs and potentially lower immunogenicity than viral vectors. None are yet in clinical use for CRISPR, but gold nanoparticle systems have shown promise in preclinical models for muscular dystrophy and other conditions where electroporation is impractical.
Extracellular Vesicles
Extracellular vesicles (EVs) — tiny membrane-bound particles naturally secreted by cells — are an emerging delivery platform. They can be engineered to carry CRISPR RNP complexes and naturally traffic to specific cell types. EVs have several theoretical advantages: they are inherently biocompatible (derived from cells), can cross biological barriers including the blood-brain barrier, and do not trigger the same immune responses as synthetic nanoparticles or viral vectors. The field is early but moving quickly.
Section 7 — Ex Vivo vs In Vivo: Two Fundamentally Different Strategies
The single most important strategic decision in CRISPR therapeutics is whether to edit cells outside the body and return them (ex vivo), or to deliver the editing machinery directly to cells inside a living patient (in vivo). These two approaches have completely different risk profiles, delivery requirements, and clinical applicability.
Cells are removed from the patient, edited in a laboratory dish, quality-controlled, then reinfused. The editing happens outside the body. You can fully characterise the edited cells before they re-enter the patient — sequencing to confirm on-target editing, testing for off-target cuts, checking for chromosomal rearrangements.
- Blood diseases (sickle cell, thalassaemia, leukaemia)
- Immune cell therapies (T cells, NK cells)
- Any disease where target cells can be extracted and reinfused
- Only works for cells that can be harvested and reinfused
- Requires myeloablative conditioning (chemotherapy) to create space for edited cells
- Expensive and logistically complex manufacturing process
CRISPR components are delivered directly into the patient. The editing happens inside the body, in the target tissue. No cell extraction, no manufacturing, no reinfusion. A single injection could potentially edit billions of cells simultaneously. This is the Holy Grail of CRISPR therapeutics — and the hardest engineering challenge.
- Liver diseases (ATTR, haemophilia, hypercholesterolaemia)
- Eye diseases (Leber congenital amaurosis, Stargardt)
- Muscle diseases (Duchenne MD) — still in development
- Any disease where target cells cannot be extracted
- Cannot characterise edits before they occur in patient
- Immune response to delivery vehicle and Cas9
- Difficult to achieve sufficient editing in deep tissues
- Off-target edits cannot be screened out before treatment
Section 8 — Delivery for Specific Tissues: Liver, Eye, Brain, Lung
Liver: The Easiest Target
The liver is currently the most accessible organ for in vivo CRISPR editing. Hepatocytes (liver cells) are highly fenestrated (have gaps in the blood vessel walls that allow large particles to pass through), avidly take up ApoE-decorated LNPs, and are the site of production for many disease-causing proteins (transthyretin in ATTR amyloidosis, PCSK9 in hypercholesterolaemia, coagulation factors in haemophilia). Multiple clinical programmes are actively editing the liver via intravenous LNP delivery, with Intellia Therapeutics’ NTLA-2001 (targeting transthyretin) showing greater than 90% protein knockdown after a single dose in Phase 1 trials.
Eye: An Immunologically Privileged Site
The eye is an attractive target for in vivo CRISPR for two reasons: it is immunologically privileged (reduced immune surveillance, lower risk of inflammatory reactions to viral vectors), and it requires very small doses of editing reagent (only the cells of the retina need to be edited, and the total volume is tiny). Editas Medicine’s EDIT-101, delivered by subretinal injection of AAV5 carrying SaCas9, is in clinical trials for Leber congenital amaurosis type 10 — a rare inherited blindness caused by a mutation in the CEP290 gene.
Brain: The Hardest Target
The brain presents the greatest delivery challenge of any organ. The blood-brain barrier (BBB) — a specialised layer of tightly joined endothelial cells lining brain blood vessels — prevents virtually all large molecules, including CRISPR components, from crossing from the bloodstream into brain tissue. Strategies under investigation include: direct intracranial injection (effective but invasive), intrathecal injection (into cerebrospinal fluid), engineered AAV variants that cross the BBB after intravenous injection (AAV-PHP.B, AAV-PHP.eB), and receptor-targeted LNPs. No CRISPR CNS therapy has yet reached clinical trials, but the preclinical data for conditions like Huntington’s disease and ALS is encouraging.
Lung: The Inhaled Delivery Frontier
The lung is an attractive target for cystic fibrosis (caused by mutations in the CFTR gene in airway epithelial cells) and for certain cancers. Inhalable LNP formulations that can be nebulised and inhaled are in development. The challenge is getting sufficient delivery to the bronchial epithelium, which is covered in mucus that traps particles. Engineered LNPs with mucus-penetrating PEG coatings and optimised particle sizes are showing promise in animal models.
Section 9 — How Casgevy Solved the Delivery Problem for Sickle Cell Disease
Casgevy (exa-cel, developed by Vertex Pharmaceuticals and CRISPR Therapeutics) is the world’s first approved CRISPR medicine, and its delivery strategy is a masterclass in choosing the right approach for the right disease. Understanding exactly how it works reveals why the ex vivo + electroporation approach was chosen and what the clinical consequences of that choice are.
The Surgeon and the Locked Fortress
Imagine a vast fortress with millions of rooms — each room a cell in the human body. Inside each room, in a vault at the back, is the genome. You are a surgeon. You have the perfect instruments: a precise molecular scalpel (Cas9) and a GPS map (the guide RNA) pointing to the exact location in the vault that needs correction. But there is one enormous problem. The fortress has walls, gates, guards, and locked doors at every level. You cannot simply walk in.
The first decision you face: what form to carry your instruments in? You could write the blueprint for your scalpel on paper (DNA plasmid) and have factory workers inside each room build it once you get the blueprint in. That takes time — hours or days — and the blueprint might accidentally get filed permanently in the room’s archive (genome integration). Or you could carry a voice recording of the instructions (mRNA) that workers act on immediately and discard afterward — faster, no permanent record. Or you could simply bring the fully assembled scalpel itself (RNP) — ready to work the moment it arrives, and gone within hours.
Now: how do you get in? The AAV approach is like hiring a trained carrier pigeon — a tiny virus stripped of anything harmful, engineered to fly precisely to the rooms you need. Different pigeons (serotypes) know different buildings: AAV9 knows how to find neurons, AAV8 finds liver rooms, AAV2 flies to the back of the eye. The pigeon is small, trusted, and knows the building well. Its one limitation: it can only carry so much. If your instruments weigh more than the pigeon can carry, you must find smaller instruments, split the load, or choose a different carrier.
The LNP approach is like a delivery van — a fat bubble that can carry far more cargo than a pigeon, packaged in a neutral-looking outer coat so the security guards (immune system) don’t stop it. The van drives through the bloodstream and, because it picks up a special flag (ApoE protein), the liver’s loading docks automatically accept it. Inside the liver, the van parks, the flag signals “authorised delivery,” and the gates open. The catch: nearly every van ends up at the liver docks. To reach rooms in the brain or lungs, you need to redesign the van entirely, put new address labels on it, or find a different vehicle.
Electroporation is different. You don’t travel to the fortress at all. Instead, you send a message: “Send me the rooms that need fixing.” The rooms — blood stem cells — are collected and brought to your laboratory. There, you use a precise electrical shock to momentarily stun the security system: for a fraction of a second, the cell membrane’s gates pop open. You push the instruments straight through. The gates close. The instruments are inside. You correct the vault. Then you send the rooms back. This is ex vivo editing — the approach that gave us Casgevy.
Casgevy’s story illustrates the whole strategy perfectly. The disease: sickle cell. The target: blood stem cells, which live in the bone marrow but can be coaxed out into the blood and collected. The correction: a CRISPR edit to reactivate a silenced gene (fetal haemoglobin) that compensates for the broken one. The stem cells are extracted, brought to the laboratory, electroporated with Cas9 RNP, and characterised exhaustively before any goes back into the patient. No guessing, no hoping — every returned cell is known to be edited correctly. The patient receives myeloablative chemotherapy to clear the old, broken bone marrow, and the corrected stem cells are infused. They take root. They multiply. Every blood cell they make produces functional haemoglobin. The disease, in most patients, is functionally cured.
The fortress analogy makes one thing clear: delivery is not a footnote to the CRISPR story. It is the story. You can have the most precise scalpel in history. But if you cannot get it into the room, the surgery never happens. The next generation of CRISPR medicines will be defined not by the guide RNA or the Cas9 variant, but by whoever solves the delivery problem for the remaining hard tissues — muscle, brain, lung — that are still, today, locked fortresses waiting for the right key.
References & Further Reading
- Naldini (2015) — Gene therapy returns to centre stage. Nature 526:351. — Comprehensive overview of viral and non-viral gene delivery strategies.
- Frangoul et al. (2021) — CRISPR-Cas9 Gene Editing for Sickle Cell Disease and Beta-Thalassemia. NEJM 384:252. — The pivotal clinical trial data for Casgevy.
- Finn et al. (2018) — A Single Administration of CRISPR/Cas9 Lipid Nanoparticles Achieves Robust and Persistent In Vivo Genome Editing. Cell Reports 22:2227. — Landmark LNP delivery for CRISPR in vivo.
- Anzalone et al. (2022) — Programmable deletion, replacement, integration and inversion of large DNA sequences with twin prime editing. Nature Biotechnology. — Advances in precision in vivo delivery strategies.
- Shirley et al. (2020) — Immune Responses to Viral Gene Therapy Vectors. Molecular Therapy 28:709. — Comprehensive review of immune barriers to AAV delivery.
- Semple et al. (2010) — Rational design of cationic lipids for siRNA delivery. Nature Biotechnology 28:172. — The original ionisable LNP paper that underpins all modern LNP delivery.
- Addgene CRISPR Delivery Guide — addgene.org/guides/crispr — Practical delivery selection guide with protocol recommendations by cell type.
- Delivery is the hardest problem in CRISPR therapeutics. A perfect guide RNA and Cas9 protein achieve nothing if they never reach the target cell. The cell membrane, endosomes, cytoplasmic nucleases, the nuclear envelope, and immune recognition all must be overcome.
- Choose the molecular form first. DNA (persistent, integration risk), mRNA (transient, no integration), or RNP (fastest, lowest off-target, best for therapeutic use). RNP is the gold standard for ex vivo editing.
- AAV is the in vivo gold standard — but has a size limit. 4.7 kb packaging limit means SpCas9 barely fits alone. Use SaCas9, split AAV, or LNPs for liver. Serotype selection determines tissue tropism. Pre-existing immunity is a critical clinical concern.
- LNPs go to the liver because of ApoE. Ionisable lipids enable endosomal escape. Excellent for liver diseases but tissue-specific targeting beyond liver requires engineering. SORT-LNPs and targeted LNPs are in development.
- Electroporation is the gold standard for ex vivo editing. >90% efficiency in haematopoietic stem cells. Used in Casgevy. Requires cells to be extracted, edited, and reinfused. Cannot be used in vivo.
- Ex vivo gives safety; in vivo gives scale. Ex vivo allows complete quality control before returning cells to the patient. In vivo reaches tissues that cannot be extracted. The therapeutic choice of strategy depends entirely on which cells need to be edited.
- Casgevy proved it works. 29/29 sickle cell patients free of severe pain crises for 12+ months post-treatment. The ex vivo + electroporation + RNP strategy delivered the first CRISPR cure. This is the blueprint every subsequent blood disease programme is following.