Programmable nucleases, from bacterial immunity to approved medicine
A working reference on CRISPR-based editing: how targeting and repair actually determine outcome, what the current editor classes can and cannot do, and where the clinical evidence now stands.
NGG PAM; without the PAM the enzyme does not cleave, which is how the system avoids destroying the bacterium's own CRISPR array.Targeting and repair
Where the edit happens is set by the guide. What the edit is gets decided by the cell.
CRISPR-Cas9 derives from an adaptive immune system in bacteria and archaea, in which fragments of past viral genomes stored in a CRISPR array are transcribed into guides that direct Cas nucleases against matching sequences. The therapeutic version reduces this to two parts: a single guide RNA carrying a ~20-nucleotide spacer, and Cas9.
Recognition is a two-factor check. Cas9 first interrogates the DNA for a protospacer adjacent motif, then unwinds the adjacent duplex and tests for base pairing with the spacer. Mismatches in the PAM-proximal seed region are strongly destabilizing; mismatches at the distal end are often tolerated, which is the structural origin of most off-target editing.
PAM scanning
Cas9 samples the genome for
NGGsites through three-dimensional collisions rather than reading the sequence linearly.R-loop formation
The duplex is unwound directionally from the PAM and the spacer pairs with the target strand, displacing the non-target strand.
Cleavage
With a full R-loop formed, the HNH domain cuts the target strand and RuvC cuts the non-target strand, giving a predominantly blunt break 3 bp from the PAM.
Repair
The cell resolves the break. This step, not the nuclease, determines the edit.
Competing repair pathways
Non-homologous end joining
Ligates the ends directly, often with small insertions or deletions. Active throughout the cell cycle and dominant in non-dividing cells. Useful for knockouts via frameshift; unsuitable for correcting a specific mutation.
Outcome: indels, frameshift, gene disruptionHomology-directed repair
Copies from a supplied donor template, permitting precise sequence replacement. Largely restricted to S/G2 phase and inefficient in the post-mitotic cells most relevant to many diseases.
Outcome: templated correction or insertionDouble-strand breaks carry their own risks
Beyond small indels, break repair can produce large deletions, chromosomal translocations, and chromothripsis, and can activate p53. These findings motivated the move toward editors that avoid a full double-strand break.
Guide design
Paste a sequence to enumerate candidate protospacers on both strands.
This tool scans for PAM-adjacent spacers, reports GC content, and flags composition features that commonly degrade performance. It runs entirely in your browser and performs no genome-wide specificity search, so it complements rather than replaces alignment-based design tools.
Example sequence: a fragment of the human HBB coding region, the gene behind sickle cell disease and beta-thalassemia.
Editor toolbox
Nucleases, nickase fusions, and catalytically dead effectors solve different problems.
| Platform | Lesion introduced | Edits possible | Cargo | Main constraint |
|---|---|---|---|---|
| Cas9 nuclease | Blunt double-strand break | Knockout, HDR knock-in, large deletion | ~4.2 kb | Repair outcome is heterogeneous; DSB-associated rearrangements |
| Cas12a (Cpf1) | Staggered break, 5′ T-rich PAM | Knockout, multiplexed targeting from one transcript | ~3.9 kb | Requires T-rich target sites |
| Base editors (CBE / ABE) | Nick only | C→T and A→G transitions within an editing window | ~5.2 kb | Bystander edits; cannot make transversions or indels |
| Prime editors | Nick only | All twelve substitutions, small insertions and deletions | ~6.3 kb | Efficiency varies by locus; pegRNA design is demanding |
| Epigenetic editors (dCas9 fusions) | None — no DNA cut | Heritable silencing or activation without sequence change | Varies, often large | Durability and reversibility still being characterized |
| Cas13d / CasRx | None — targets RNA | Transcript knockdown, splice modulation, RNA editing | ~2.9 kb | Effect is transient; collateral cleavage must be excluded |
Base editing and prime editing both use a Cas9 nickase, so neither depends on a double-strand break or a donor template. A base editor fuses a deaminase that converts a single base within a narrow window; a prime editor fuses a reverse transcriptase and writes new sequence from an extended guide called a pegRNA, installing the edit at or downstream of the nick. That reach is why prime editing covers substitutions and small indels that base editors cannot.
RNA targeting
Cas13d — the effector this site is named for — leaves the genome untouched.
Not every CRISPR effector acts on DNA. Type VI systems use Cas13 enzymes, which bind and cleave single-stranded RNA through HEPN nuclease domains. Cas13d from Ruminococcus flavefaciens, commonly called CasRx, is unusually compact at roughly 930 amino acids, small enough to package with its guide array into a single AAV vector.
The therapeutic logic differs from genome editing. Because the target is the transcript rather than the gene, the effect is dose-dependent and reversible, and no permanent change is written into the genome. That suits gain-of-function and dominant-negative disease, repeat-expansion transcripts, and allele-selective knockdown that spares the healthy copy. It also avoids double-strand-break toxicity entirely.
The trade-offs are durability and a mechanistic caveat. Once activated by its target, Cas13 can cleave bystander RNAs non-specifically in some systems. The extent of this collateral activity in mammalian cells has been contested and depends heavily on target abundance and expression level, so any therapeutic claim needs collateral activity measured directly rather than assumed absent.
No genomic edit
Nothing heritable is written, which removes germline and insertional concerns.
Reversible
Knockdown decays as the effector clears, allowing dose titration.
Compact
Small enough for single-vector AAV packaging alongside multiplexed guides.
Delivery
The editor is rarely the bottleneck. Getting it into the right cells usually is.
| Modality | Exposure window | Tropism | Notes |
|---|---|---|---|
| Ex vivo electroporation of RNP | Minutes | Whatever is harvested | Short protein half-life limits off-target exposure; basis of the approved hematopoietic therapies |
| Lipid nanoparticle, intravenous | Hours to days | Strongly hepatic by default | Non-integrating and redosable; enabled rapid bespoke in vivo therapy |
| Adeno-associated virus | Months, persistent expression | Serotype-dependent; reaches CNS, retina, muscle | ~4.7 kb ceiling; prolonged expression increases off-target opportunity; pre-existing immunity |
| Direct local injection | Variable | Immune-privileged compartments such as the eye | Low systemic exposure; only viable for accessible tissue |
Why the liver came first
Intravenous lipid nanoparticles accumulate in hepatocytes through apolipoprotein-mediated uptake. That is why in vivo editing programs cluster around liver-expressed targets — transthyretin, PCSK9, alpha-1 antitrypsin — while extrahepatic delivery remains the field's central unsolved problem.
Specificity assessment
Off-target claims are only as strong as the assay used to make them.
No single method establishes specificity. In silico prediction is cheap but misses sites with bulges or non-canonical PAMs. Biochemical methods on naked DNA are sensitive and unbiased but ignore chromatin, so they over-call. Cell-based methods capture real chromatin context but are less sensitive. Regulatory packages generally combine an unbiased discovery method with deep targeted sequencing in the actual therapeutic cell type, across multiple donors.
| Approach | Basis | Strength | Blind spot |
|---|---|---|---|
| In silico homology search | Reference genome mismatch enumeration | Instant, exhaustive over the reference | Bulges, variant alleles, non-reference genomes |
| Cell-free biochemical cleavage | Purified genomic DNA digested in vitro | Unbiased and highly sensitive | Ignores chromatin; many false positives |
| Break capture in cells | Tagging or marking double-strand breaks in situ | Native chromatin context | Lower sensitivity for rare events |
| Targeted deep sequencing | Amplicon sequencing of nominated sites | Quantitative confirmation below 0.1% | Only sees sites already nominated |
| Long-read or karyotypic analysis | Structural variant detection | Finds large deletions and translocations | Depth and cost |
Two design levers reduce the problem at source. High-fidelity Cas9 variants weaken non-specific protein–DNA contacts so that mismatched R-loops fail to trigger cleavage, and delivering pre-formed ribonucleoprotein rather than DNA or mRNA shortens the window during which the nuclease is present at all.
Clinical status
Approved therapies, active programs, and the fastest translation on record.
| Indication | Strategy | Route | Status |
|---|---|---|---|
| Sickle cell disease | Cas9 disruption of the BCL11A erythroid enhancer to de-repress fetal hemoglobin | Ex vivo, autologous HSPC | Approved |
| Transfusion-dependent beta-thalassemia | Same enhancer-disruption approach | Ex vivo, autologous HSPC | Approved |
| Transthyretin amyloidosis | Knockout of hepatic TTR to reduce circulating amyloidogenic protein | In vivo, LNP | Late phase |
| Hereditary angioedema | Knockout of KLKB1 to suppress bradykinin generation | In vivo, LNP | Late phase |
| Refractory hypercholesterolemia | Base editing of PCSK9 to lower LDL cholesterol durably | In vivo, LNP | Early phase |
| Alpha-1 antitrypsin deficiency | Base or prime correction of the pathogenic SERPINA1 allele | In vivo, LNP | Early phase |
| X-linked severe combined immunodeficiency | Base editing of IL2RG in autologous stem cells | Ex vivo | Early phase |
| CPS1 deficiency | Bespoke in vivo base editing designed for one infant's private mutation | In vivo, LNP | n-of-1 |
| Oncology, CAR-T | Multiplexed editing for allogeneic, off-the-shelf cell products | Ex vivo | Multiple trials |
Trajectory
- 1987
CRISPR arrays observed
Unexplained repeat structures are noted in the E. coli genome, with no function assigned for two decades.
- 2007
Adaptive immunity demonstrated
Phage challenge of Streptococcus thermophilus shows the locus confers acquired resistance.
- 2012–2013
Programmable cleavage, then mammalian editing
Cas9 is shown to be RNA-guided and retargetable, and is applied in human cells.
- 2016–2019
Precision without cutting
Base editors and then prime editors decouple editing from the double-strand break.
- 2021
In vivo editing in humans
Systemic LNP delivery produces durable knockdown of a hepatic target in patients.
- 2023–2024
First approvals
A CRISPR-Cas9 therapy is authorized for sickle cell disease, then for beta-thalassemia.
- 2025 →
Bespoke and repeatable
A personalized in vivo base-editing therapy is designed, authorized, and administered to a single infant in roughly six months, pointing toward a platform-and-payload regulatory model rather than one drug per disease.
Open questions
Extrahepatic delivery, long-term durability of edited stem cell grafts, clonal dynamics after editing, the cost and logistics of autologous manufacturing, access in the regions carrying the highest sickle cell burden, and governance of heritable germline modification — which remains broadly prohibited and is scientifically distinct from the somatic work described here.
Sources and further reading
Primary literature and maintained trial registries.
- CRISPR Medicine News clinical trials registerCurated, continuously updated listing of gene-editing trials across Cas nucleases, base and prime editors, and RNA editors.
- CRISPR clinical trials: annual updateInnovative Genomics Institute. Indication-by-indication summary of active programs.
- A primer on prime: prime editing from advances to first-in-human trialMolecular Therapy. Mechanism, pegRNA architecture, and remaining efficiency and delivery challenges.
- Advances in CRISPR base editingJournal of Cellular and Molecular Medicine. Editor evolution, bystander editing, and therapeutic application.
- Advancing CRISPR genome editing into gene therapy clinical trialsReview covering ZFNs and TALENs through to approved CRISPR therapy, with emphasis on translational obstacles.
- Genomics reference materialNational Human Genome Research Institute. Background suitable for non-specialist collaborators and patients.