CRISPR/Cas9: mechanism, repair, and delivery

CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) is, biologically, an adaptive immune system that many bacteria and archaea use against viruses. The bacterial machinery stores fragments of viral DNA from previous infections in a 'CRISPR array' in the bacterial genome. When the same virus invades again, the stored fragments are transcribed into guide RNAs that program a nuclease (Cas9 in the simplest system) to find and cut the viral DNA.

The structural insight that produced the genome-editing field: the same machinery can be reprogrammed to target any DNA sequence by changing the guide RNA. If you can synthesize a guide RNA that matches a sequence you want to edit, Cas9 will find that sequence in any cell containing the machinery and cut it.

This was demonstrated in test tubes and bacteria in 2012 (Jinek, Charpentier, Doudna) and adapted for mammalian cells in 2013 (multiple groups including Zhang, Church). The Nobel Prize in Chemistry 2020 was awarded to Charpentier and Doudna for the discovery. The structural simplicity — one protein, one programmable guide — explains why CRISPR rapidly displaced earlier genome-editing technologies (zinc-finger nucleases, TALENs) that required custom protein engineering for each target.

The minimal CRISPR/Cas9 editing system consists of two components.

Cas9. A large nuclease protein (~1400 amino acids in the most-used variant, from Streptococcus pyogenes). Cas9 has two nuclease domains, RuvC and HNH, each of which cuts one strand of DNA. Together they produce a double-strand break.

Single-guide RNA (sgRNA). An engineered RNA of about 100 nucleotides that combines two natural CRISPR RNAs into one. The sgRNA has two regions:

• A spacer sequence of 20 nucleotides that matches the target DNA by base pairing. This is the programmable part.
• A scaffold sequence that binds Cas9 and positions it for cleavage.

The sgRNA-Cas9 complex (the ribonucleoprotein, or RNP) is the working editing unit. It scans DNA looking for a sequence complementary to the spacer.

The activity of the complex follows a strict rule: the target DNA must end (on its 3' side) in a short protospacer adjacent motif (PAM), specifically NGG for S. pyogenes Cas9. The PAM is not part of the guide; it is required for Cas9 to bind the target at all. The PAM constraint limits which sequences in the genome are targetable by any given Cas9 variant; engineered variants with different PAM requirements (NG, NGN, others) have been developed to widen the targetable space.

Cas9 holds the guide RNA, scans for a matching DNA sequence with the right PAM, and cuts.

Cas9's cut produces a double-strand break (DSB) in the genomic DNA. DSBs are dangerous to cells — they can lead to chromosome translocations or cell death if not repaired. Cells respond to DSBs by repair pathways, and which pathway acts determines whether the edit ends up as a disruption or a correction.

Two major pathways:

• Non-homologous end joining (NHEJ). The cell's default DSB repair. The cell rejoins the two broken ends directly, often with small insertions or deletions (indels) at the join. NHEJ is fast and active in all phases of the cell cycle. It is error-prone: the rejoined sequence differs from the original by a few bases.
• Homology-directed repair (HDR). A slower pathway that uses a homologous DNA template (the sister chromatid in dividing cells, or a synthetic template supplied with the edit) to repair the break precisely. HDR is active mainly in S and G2 phases (when sister chromatids are available) and is much less efficient than NHEJ in most cell types.

The practical consequences:

• For knocking out a gene, NHEJ is sufficient. Cas9 cuts within the gene; NHEJ creates indels; the resulting frameshifts or premature stops disrupt the protein. NHEJ-mediated knockout is the highest-efficiency CRISPR application.
• For correcting a specific mutation, HDR is needed (or a base/prime editor — next lesson). Supply a template DNA matching the corrected sequence; HDR uses it to rewrite the break with the corrected version. HDR-mediated correction is much less efficient — often 1–10% in dividing cells, near zero in non-dividing cells.

This efficiency asymmetry — knockout easy, correction hard — has shaped the early clinical applications of CRISPR. Most approved or near-approval CRISPR therapies use the easier mode.

Cas9 finds its target by base-pairing the sgRNA spacer with the DNA. Real-world Cas9 is imperfect at this matching: spacers tolerate some mismatches and still cut. A guide designed to target one site can also cut at unrelated sites elsewhere in the genome that happen to be similar.

Off-target cuts can have consequences:

• Loss-of-function mutations in unrelated genes. An off-target cut in a tumor-suppressor gene, for example, could be carcinogenic.
• Chromosomal rearrangements. Simultaneous on-target and off-target cuts can join chromosomes incorrectly.
• Larger structural variants. Deletions, inversions, and translocations between any two simultaneous cut sites.

Quantifying off-target effects:

• GUIDE-seq, CIRCLE-seq, DISCOVER-seq, others — sequencing-based assays that identify cut sites in cells or in vitro.
• Genome-wide off-target prediction tools identify potential off-target sites computationally based on sequence similarity, but predictions miss some real off-targets and overcall others.
• Whole-genome sequencing of edited cells detects mutations directly but is expensive at the per-cell scale needed for clinical evaluation.

Reducing off-target effects has been an active engineering area. Strategies include:

• High-fidelity Cas9 variants (eSpCas9, SpCas9-HF1, HiFi-Cas9) with engineered amino-acid substitutions that reduce off-target activity.
• Cas9 nickases (one of the two nuclease domains inactivated) that cut only one strand; pairs of nickases must engage to produce a DSB, sharply reducing off-target probability.
• Truncated guides of less than 20 nucleotides, which are less tolerant of mismatches.
• Time-limited delivery — short-lived Cas9 (as RNP or mRNA) cuts on a tight schedule and degrades, reducing time for off-target events.

CRISPR's biological mechanism is well understood; delivery — getting the editing machinery into the right cells in the right organ at the right time — is the harder engineering problem for clinical translation.

Two categories of delivery dominate.

Ex vivo. Take cells out of the patient (e.g., hematopoietic stem cells from bone marrow or peripheral blood), edit them in the laboratory, then return the edited cells to the patient. Delivery in the lab is straightforward — electroporation of the Cas9-sgRNA RNP into cells achieves high efficiency. The clinical challenges are conditioning the patient to accept the edited cells (often myeloablative conditioning for hematopoietic edits) and managing the production-and-return logistics.

In vivo. Deliver the editing machinery directly to cells in the body. Delivery vehicles include:

• Adeno-associated virus (AAV) — small viral capsids carrying DNA encoding Cas9 and sgRNA. Targets specific tissues based on capsid serotype. Expression persists for months to years. Limitations: cargo size limit (~4.7 kb makes packaging full-length Cas9 tight), prior immunity in some patients, integration concerns.
• Lentivirus — larger DNA cargo, but integrates into the host genome (carries its own risks).
• Lipid nanoparticles (LNPs) — synthetic lipid vesicles carrying mRNA encoding Cas9 plus sgRNA. Transient expression, no integration, no viral immunity, but limited tissue targeting (mostly liver after IV delivery).
• Direct injection — for accessible tissues (eye, skin, muscle).

In-vivo delivery to most tissues remains challenging. Liver-targeted LNP delivery is the most developed; CNS, muscle, and many epithelial tissues are active research areas. Many CRISPR therapies in clinical trials are limited to diseases where the affected cells are accessible ex vivo or where AAV/LNP delivery to the relevant tissue has been demonstrated.

Casgevy (exa-cel) is the first regulatory-approved CRISPR therapy. Approved by the UK MHRA in November 2023, the US FDA in December 2023, and the European EMA shortly after, it treats sickle-cell disease and transfusion-dependent beta-thalassemia.

The therapy's structure illustrates the principles in this lesson.

• Disease target. Both sickle-cell disease and beta-thalassemia involve mutations in the adult beta-globin gene that disrupt hemoglobin. Patients normally have fetal hemoglobin during gestation, but a developmental switch suppresses fetal hemoglobin and turns on adult hemoglobin after birth.
• Edit strategy. Rather than correct the disease mutation (HDR-difficult), Casgevy uses NHEJ to disrupt the gene encoding the fetal-hemoglobin repressor (BCL11A's erythroid enhancer). With the repressor silenced, patients re-express fetal hemoglobin, which compensates for the disease.
• Delivery. Ex vivo. Patient's hematopoietic stem cells are collected, edited with Cas9 + sgRNA targeting the BCL11A enhancer, and re-infused after myeloablative conditioning.
• Outcome. In trial data, the great majority of treated patients with sickle-cell disease have not had vaso-occlusive crises in the follow-up period; transfusion-dependent beta-thalassemia patients have become transfusion-independent.
• Costs and access. List price about $2.2 million per patient. Treatment requires specialized centers, multi-month protocols, and managed-care complexity.

The Casgevy translation shows how the first CRISPR clinical successes have favored ex vivo editing of accessible stem cells with NHEJ-mediated disruption of a regulatory element, in a disease with a clear mechanism. The principles of in vivo delivery and HDR-based correction are still being scaled.

Three structural points for the cursus.

• The editing machinery is conceptually simple. Cas9 + sgRNA + PAM rule. The complexity is in the cellular response (NHEJ vs HDR), the off-target landscape, and the delivery problem.
• NHEJ is easier than HDR. Most current clinical applications exploit this — disruption of a regulatory element or gene that yields therapeutic benefit, rather than precise correction of the disease-causing mutation.
• Delivery, not editing, is the bottleneck for many applications. Reaching the right cells in the right tissues with the right timing is the engineering problem that determines which diseases become tractable when.

The next lesson extends the editing framework beyond cutting: base editors and prime editors that achieve precise single-base or short-sequence changes without making double-strand breaks at all, addressing the off-target and HDR-efficiency limitations of standard CRISPR/Cas9.