CRISPR-Cas9 Genome Editing

CRISPR stands for Clustered Regularly Interspaced Short Palindromic Repeats, a family of DNA sequences found in bacterial and archaeal genomes consisting of short repeated sequences separated by unique "spacer" sequences derived from past viral or plasmid invaders. Adjacent to the CRISPR array sit cas (CRISPR-associated) genes encoding nucleases. Together, the CRISPR array and Cas proteins form a heritable, sequence-specific immune memory: when a virus attacks a bacterium a second time, RNA transcribed from the matching spacer guides a Cas nuclease to cut the invading DNA before it can replicate. The type II CRISPR system from Streptococcus pyogenes uses a single multidomain protein, Cas9, together with two small RNAs: crRNA (carrying the sequence-specific spacer) and tracrRNA (a scaffold RNA that base-pairs with crRNA and recruits Cas9). For laboratory use these two RNAs are fused into a single synthetic single-guide RNA (sgRNA), roughly 100 nucleotides long, with a customizable 20-nucleotide "spacer" at its 5′ end that determines what the complex targets. This engineering simplification — reducing a three-component natural system to two components (Cas9 protein + one RNA) — is what made CRISPR practical as a general-purpose tool. Critically, Cas9 does not simply search for any matching 20-mer. It requires a short adjacent DNA motif called the protospacer-adjacent motif (PAM) — for S. pyogenes Cas9, the sequence 5′-NGG-3′ immediately 3′ of the target on the non-target strand. The PAM requirement is what originally let bacteria distinguish "self" (their own CRISPR array, which lacks a PAM next to its spacers) from "non-self" invader DNA, preventing autoimmune self-cleavage. In genome editing, the PAM constrains which genomic sites are targetable — roughly one every 8–12 bp in mammalian genomes for NGG — and different Cas9 orthologs or engineered variants recognize different PAMs, expanding the targetable sequence space.

Step 1 — Guide RNA loading. Cas9 protein binds the sgRNA (or the crRNA:tracrRNA duplex in the natural system), undergoing a conformational change that pre-organizes the RNA's spacer sequence for base pairing and exposes the PAM-interacting domain.

Step 2 — PAM scanning. The Cas9–RNA complex diffuses along DNA, transiently sampling sequences until its PAM-interacting domain detects a 5′-NGG-3′ motif. This scanning step, not base-pairing, is the actual rate-limiting search process — Cas9 essentially checks PAMs first and only then tests whether the adjacent sequence matches the guide.

Step 3 — R-loop formation. Once a PAM is found, Cas9 locally unwinds the adjacent DNA duplex and the guide RNA invades the DNA, base-pairing with the target (complementary) strand to form an RNA:DNA heteroduplex, displacing the non-target strand as a single-stranded loop — an R-loop. Pairing must be near-perfect in the seed sequence (the ~10–12 nucleotides immediately proximal to the PAM); mismatches here abolish cleavage, while mismatches distal from the PAM are more tolerated, a major source of off-target activity.

Step 4 — Cleavage. Correct R-loop formation triggers a conformational change activating Cas9's two nuclease domains: HNH, which cuts the target (RNA-paired) strand, and RuvC, which cuts the non-target strand. Each domain nicks one strand 3 bp upstream (5′ side) of the PAM, generating a blunt-ended double-strand break (DSB).

Step 5 — Repair by the host cell. Cas9 itself does not "edit" anything — it only cuts. Editing outcomes are entirely determined by which DNA repair pathway seals the break:

NHEJ directly re-ligates the blunt ends but frequently loses or gains a few nucleotides at the junction, especially after repeated cutting (since a perfectly re-ligated site is simply re-cut by Cas9 until an indel disrupts the PAM or seed sequence). A frameshift indel in a coding exon typically destroys the reading frame, producing a null allele — this is the basis of CRISPR knockout screens, a direct extension of classical loss-of-function logic but achieved by cutting rather than by gene replacement. HDR instead uses a supplied donor template — a plasmid or single-stranded oligonucleotide bearing the desired sequence change flanked by sequences homologous to the cut site — and the cell's homologous recombination machinery (RAD51-dependent strand invasion, as in meiotic recombination) copies the template across the break, introducing precise point mutations, tags, or insertions. Because HDR requires a sister chromatid and the S/G2-phase-specific recombination machinery, its efficiency is low (often single-digit to ~20% of edited alleles) and can be increased by synchronizing cells or by chemically suppressing competing NHEJ.

Step 6 — Repair fate diverges. After the break, two competing outcomes race against each other in every treated cell — most cells default to error-prone NHEJ unless a donor template and cell-cycle conditions favor HDR.

Discovery of the repeats (Ishino et al., 1987). While sequencing the iap gene region in E. coli, Yoshizumi Ishino and colleagues noticed an unusual array of 29-bp direct repeats separated by variable 32-bp spacers downstream of the gene. They could not explain the function and noted it explicitly as a mystery in their paper — the first documented sighting of what would later be named CRISPR.

Spacers match viral sequences (Mojica, Bolotin, Pourcel; 2005). Francisco Mojica's bioinformatic survey across many bacterial and archaeal genomes showed that CRISPR spacers were not random — they matched sequences from bacteriophages and plasmids, almost never from the host's own genome. Mojica proposed that CRISPR functions as an adaptive immune system, a hypothesis published in 2005 after being rejected by several journals.

Functional immunity demonstrated (Barrangou et al., 2007). Working at Danisco on the dairy bacterium Streptococcus thermophilus, Rodolphe Barrangou and Philippe Horvath showed experimentally that bacteria surviving phage infection had acquired new spacers matching the phage genome, and that adding or removing specific spacers respectively increased or eliminated resistance to that phage. This was the first direct experimental proof that CRISPR spacers confer sequence-specific, heritable immunity — not merely correlation, but causation, since deleting a spacer removed resistance and adding one conferred it.

Cas9 identified as the effector nuclease and tracrRNA discovered (Deltcheva et al., 2011). Emmanuelle Charpentier's group identified tracrRNA and showed it was required for crRNA maturation and for Cas9 function in S. pyogenes, clarifying the two-RNA architecture of the type II system.

In vitro reconstitution of a programmable nuclease (Jinek, Chylinski, Fonfara, Hauer, Doudna, Charpentier — Science 2012). This is the pivotal experiment. Jennifer Doudna and Emmanuelle Charpentier's teams purified Cas9 protein and combined it in a test tube with crRNA and tracrRNA, then showed the complex cut purified target DNA precisely at the site specified by the crRNA sequence, 3 bp upstream of a PAM. Crucially, they demonstrated that crRNA and tracrRNA could be fused into a single chimeric guide RNA without loss of activity — the invention that converted a natural three-part bacterial system into a two-component tool any lab could reprogram simply by changing 20 nucleotides of guide sequence. This single paper is generally credited as the founding demonstration of programmable Cas9 genome editing, and it underlay the 2020 Nobel Prize in Chemistry to Doudna and Charpentier.

Editing demonstrated in human and mammalian cells (Cong et al., Zhang lab; Mali et al., Church lab — both Science, January 2013). Independently and almost simultaneously, Feng Zhang's group and George Church's group showed that Cas9 with a chimeric sgRNA could cleave and edit genomic loci in cultured human and mouse cells, producing targeted indels via NHEJ and precise edits via HDR with a donor template, and that multiple genes could be targeted simultaneously ("multiplexing") by co-delivering several sgRNAs. This closed the gap between a biochemical demonstration and a usable genome-engineering method for eukaryotic research and, eventually, therapy.

Off-target and specificity studies (Fu et al. 2013; Hsu et al. 2013, and others). Genome-wide sequencing after Cas9 treatment revealed that mismatches distal from the PAM are tolerated more than seed-region mismatches, explaining variable off-target cleavage and motivating truncated guides, high-fidelity Cas9 variants (e.g., eSpCas9, HiFi Cas9), and paired-nickase strategies that require two adjacent cuts to reduce off-target double-strand breaks.

Example 1 — Predicting a CRISPR knockout cross in Drosophila. A researcher injects Cas9 mRNA and an sgRNA targeting exon 2 of an autosomal gene white+ (eye pigmentation) into fertilized Drosophila embryos from a true-breeding red-eyed (white+/white+) stock. Injected G0 flies are mosaic — only some germline cells carry the edit. She crosses a single G0 male to white–/white– (white-eyed) virgin females and screens 200 G1 progeny for eye color. Suppose 38 of 200 G1 flies (19%) are white-eyed. Because G0 is mosaic, this fraction reflects the proportion of germline cells that received a biallelic disrupting indel — here, effectively, the fraction of gametes carrying a loss-of-function white allele from the injected parent (since the other parent contributes only null alleles). A white-eyed G1 fly is white(edited)/white–, confirming that a single NHEJ-induced frameshift indel is sufficient to produce a fully recessive-phenotype null allele, exactly as classical loss-of-function theory predicts for a gene required for pigment deposition. Sequencing several white-eyed G1 flies' white locus shows a mix of indel sizes (−1, −4, +2 bp, etc.) at the same cut site — direct molecular evidence that NHEJ, not a single defined mutation, generated the alleles, since each fly's lesion differs even though all were cut at the identical Cas9 target site.

Example 2 — Efficiency arithmetic for a precise HDR knock-in. A lab wants to introduce a single point mutation (a disease-associated missense allele, c.1521_1523delCTT — the ΔF508 mutation studied in cystic fibrosis research models) into human iPSCs using Cas9 plus a single-stranded oligonucleotide donor. They electroporate 500,000 cells with Cas9 ribonucleoprotein (RNP), sgRNA, and donor, then genotype 96 surviving clonal colonies by targeted sequencing. Results: 58 of 96 clones (60%) show an indel from NHEJ; 12 of 96 clones (12.5%) show the precise HDR-mediated edit (of which 5 are heterozygous, 7 are homozygous); the remainder are unedited wild type. Interpreting this: the roughly 5:1 ratio of NHEJ to HDR outcomes (58:12) is typical, because NHEJ operates throughout the cell cycle and out-competes HDR, which is restricted to S/G2 phase. To recover a homozygous edited line for downstream disease modeling, the lab needs only the 7 homozygous clones — but because Cas9 continues to cut and NHEJ can act on the already-HDR-edited allele too, some "heterozygous" clones may actually carry one HDR allele and one re-cut, indel-bearing allele rather than a true wild-type allele; this must be checked by sequencing both alleles fully, not inferred from a single PCR band, before concluding the genotype. This illustrates why real CRISPR knock-in experiments report clone numbers and full-locus sequencing, not just "editing efficiency" as a single percentage.

Jennifer Doudna trained as a structural biochemist studying RNA catalysis — ribozymes and the structure of the group I intron — long before CRISPR entered her research program. In 2011, a chance conversation at a scientific conference in Puerto Rico connected her with Emmanuelle Charpentier, a microbiologist who had spent years studying the type II CRISPR system's tracrRNA and Cas9 in Streptococcus pyogenes. Doudna's structural and biochemical expertise in RNA-protein complexes combined with Charpentier's deep knowledge of the natural bacterial system to produce the 2012 Science paper reconstituting Cas9 as a programmable nuclease in vitro — the single experiment most often cited as the technology's founding moment. Doudna has since led structural studies resolving how Cas9 recognizes PAMs and forms the R-loop, work that directly explains the seed-sequence and off-target phenomena described above.

Functional genomics. Genome-wide CRISPR knockout and CRISPRi/CRISPRa screens (using catalytically "dead" Cas9, dCas9, fused to repressor or activator domains) systematically identify genes required for a phenotype — extending classical forward and reverse genetic screening to genome scale in human cell lines, something impossible with earlier single-gene knockout technologies.

Medicine. CRISPR-based ex vivo gene editing of patient hematopoietic stem cells underlies exagamglogene autotemcel (Casgevy), approved by the FDA and UK MHRA in 2023–2024 for sickle cell disease and beta-thalassemia — editing a regulatory sequence to reactivate fetal hemoglobin expression, compensating for defective adult hemoglobin. In vivo lipid-nanoparticle-delivered CRISPR (e.g., NTLA-2001) is in clinical trials for transthyretin amyloidosis, directly knocking out the disease-causing gene in patient liver cells.

Agriculture. CRISPR has produced non-transgenic crop edits (no foreign DNA remains after selection) such as disease-resistant rice, high-oleic soybeans, and browning-resistant mushrooms, several of which face lighter regulatory hurdles than classical transgenics because the final product carries no foreign DNA — only an indel or point mutation indistinguishable from natural variation.

Diagnostics. Cas12a/Cas13-based detection platforms (e.g., SHERLOCK, DETECTR) exploit collateral nuclease activity triggered by target recognition to build rapid, low-cost nucleic acid diagnostics, including for SARS-CoV-2 and other pathogens — a direct biotechnological offshoot of the same guide-RNA-targeting logic.

Conservation and ecology. Gene drive systems built on CRISPR (designed to bias inheritance toward an edited allele in a population) are being engineered and debated as tools to suppress invasive species or disease-vector mosquito populations, raising major ecological and governance questions precisely because the technology's efficiency at spreading an edit through a population is so high.

Ask the tutor: How do researchers actually choose and validate a good sgRNA sequence before running a CRISPR experiment?