How Cas9 Cuts DNA: The Complete Molecular Mechanism

Section 1 — Overview: The Two-Component System

Before diving into molecular detail, it helps to have the big picture clearly in mind. The CRISPR-Cas9 editing system has exactly two components that you deliver to a cell. Everything else — every conformational change, every nucleotide check, every phosphodiester bond broken — happens automatically as a consequence of the physics and chemistry of these two molecules finding each other and finding their target.

Component one is the Cas9 protein — a large, multi-domain enzyme (~160 kilodaltons, about 1,368 amino acids in the most commonly used variant from Streptococcus pyogenes) that does the actual DNA cutting. On its own, Cas9 is essentially inactive. It binds DNA non-specifically and weakly. It cannot cut anything.

Component two is the guide RNA — a short RNA molecule (~100 nucleotides total) that tells Cas9 exactly where to cut. It has two functional regions: a 20-nucleotide spacer sequence that base-pairs with the target DNA, and a structural scaffold region that binds to and activates Cas9. When guide RNA binds Cas9, the complex undergoes a dramatic conformational change that activates the protein and arms it for cutting.

Section 2 — Cas9 Protein Architecture: A Tour of the Domains

Understanding Cas9’s mechanism requires understanding its physical structure. Cas9 is not a simple, featureless enzyme. It is a multi-domain protein that changes its three-dimensional shape at each stage of the cutting process. The different domains have different jobs, and mutations in any one of them produce a protein with different capabilities — which is exactly how researchers engineer Cas9 variants.

Structurally, Cas9 can be divided into two main lobes that form a bilobed architecture resembling a pair of cupped hands or a crab claw. The recognition lobe (REC lobe) and the nuclease lobe (NUC lobe) are connected by an arginine-rich bridge helix and create a central channel through which the guide RNA-DNA heteroduplex (the R-loop) is threaded during the cutting reaction.

Section 3 — The Guide RNA: From Two Molecules to One

In the natural CRISPR immune system of Streptococcus pyogenes, Cas9 requires two separate RNA molecules to function. The crRNA (CRISPR RNA) contains the spacer sequence that matches the target DNA. The tracrRNA (trans-activating crRNA) base-pairs with a repeat region in the crRNA to form a duplex structure that binds and activates Cas9. Both are required. Neither works alone.

The Doudna-Charpentier lab’s key engineering breakthrough was showing that the crRNA and tracrRNA could be fused into a single RNA molecule via a short loop sequence — creating what they called the single guide RNA (sgRNA). This fusion dramatically simplified the system: instead of separately synthesising and delivering two RNA molecules, users needed only one. The sgRNA retains all the functionality of the two-piece system while being much easier to design, produce, and deliver.

Section 4 — PAM Recognition: How Cas9 Finds Its Starting Point

The genome is enormous. How does Cas9 efficiently search 3.2 billion base pairs for its target without spending an eternity at every position? The answer is the PAM sequence (Protospacer Adjacent Motif) — a short DNA sequence that Cas9 checks first before investing energy in the full guide RNA matching process.

For SpCas9 (the most widely used variant), the PAM is 5’-NGG-3’ on the non-template strand, located immediately 3’ (downstream) of the protospacer (the genomic DNA sequence that matches the guide RNA spacer). NGG means any nucleotide followed by two guanines. This sequence occurs approximately once every 8 base pairs in a random genome — frequently enough that almost any gene can be targeted, but infrequently enough to reduce the search space significantly.

The Mechanism of PAM Recognition

Cas9 scans DNA by sliding along in a 3D diffusion process — sometimes called facilitated diffusion. At each position, it briefly interrogates the DNA sequence. When it encounters an NGG, it slows down and opens the DNA duplex locally to allow the guide RNA spacer to attempt base-pairing with the exposed strand. If the PAM is wrong (not NGG), Cas9 moves on without interrogating further.

The PAM is recognised by the PAM-interacting (PI) domain of Cas9, which reads the major groove of the DNA. The two guanines of the NGG PAM make specific contacts with Arg1333 and Arg1335 residues in the PI domain. These arginine-guanine interactions are what make NGG the required PAM for SpCas9. Mutating Arg1333 or Arg1335 destroys PAM recognition and abolishes Cas9 activity.

Section 5 — R-Loop Formation: Checking the Match

Once Cas9 finds a PAM and opens the DNA locally, the guide RNA spacer must check whether the adjacent 20-nucleotide sequence actually matches the intended target. This matching process is called R-loop formation (the R stands for RNA), and it is the central specificity-checking step of the entire mechanism.

The R-loop forms as the guide RNA spacer invades the DNA double helix and displaces the non-complementary DNA strand. The 20 nucleotides of the spacer base-pair with the complementary DNA strand one by one, starting from the PAM-proximal end (nucleotides 1–3) and zipping toward the PAM-distal end (nucleotides 18–20). The displaced non-complementary DNA strand forms the “loop” of the R-loop structure.

If the spacer sequence perfectly matches the DNA, the R-loop forms stably and extends through all 20 nucleotides. This full R-loop formation is the trigger that activates Cas9 for cutting. If there are mismatches — base pairs that cannot form — the R-loop is destabilised and may fail to extend fully, leaving Cas9 inactive at that position. The stringency of this check determines Cas9 specificity.

The Directionality of R-Loop Formation

Section 6 — The Cut: HNH and RuvC in Action

The actual DNA cleavage event is carried out by Cas9’s two nuclease domains working in concert. Each domain cuts one strand of the DNA double helix, together producing a double-strand break (DSB) — a complete severance of the DNA molecule at a single defined position.

The HNH domain cuts the complementary strand — the strand that has base-paired with the guide RNA. The catalytic residue is Histidine 840 (H840). HNH is a magnesium-dependent endonuclease: it requires a Mg²⁺ ion in its active site to activate a water molecule for nucleophilic attack on the DNA phosphodiester backbone. The cut leaves a 3’-hydroxyl and a 5’-phosphate group.

The RuvC domain cuts the non-complementary (displaced) strand — the strand that was pushed out during R-loop formation. RuvC is also Mg²⁺ dependent and uses a two-metal-ion mechanism similar to many other nucleases. Its catalytic residues include Aspartate 10 (D10) and Histidine 983 (H983). The cut is made at the same position as the HNH cut: 3 base pairs upstream of the PAM sequence.

Section 7 — The Seed Region: Why Location of Mismatches Matters

Not all positions in the 20-nucleotide spacer are equal. Mismatches between the guide RNA and target DNA are much better tolerated at some positions than others. The most important concept here is the seed region: the ~12 nucleotides immediately adjacent to the PAM (positions 1–12, counting from the PAM-proximal end).

Mismatches within the seed region almost always abolish or severely reduce Cas9 cutting. This is because R-loop formation initiates at the PAM-proximal end and zips toward the PAM-distal end: if early base pairs cannot form (seed region mismatches), the R-loop stalls and cannot complete, so Cas9 never reaches the active conformation. The seed region is therefore the primary specificity checkpoint.

Mismatches in the PAM-distal region (positions 13–20) are much better tolerated. The R-loop can sometimes extend over one or even two mismatches in this region, allowing Cas9 to cut even at imperfectly matched sites. This is the primary source of CRISPR off-target effects: sites with perfect seed region matching but one or two PAM-distal mismatches can still be cut at significant frequency.

Section 8 — Cas9 Variants: Different Bacteria, Different Rules

SpCas9 from Streptococcus pyogenes was the first and is still the most widely used Cas9. But it is not the only Cas9 in nature. Dozens of bacterial species have their own CRISPR-Cas9 systems, each with different protein sizes, PAM requirements, and cutting characteristics. Understanding the alternatives matters for two reasons: different applications call for different tools, and the limitations of SpCas9 (particularly its large size and NGG-only PAM) have driven the development of alternatives.

Section 9 — Engineered Cas9: Nickases, dCas9, and What They Enable

Understanding Cas9 domain structure has enabled researchers to engineer variants with fundamentally different capabilities. By mutating the catalytic residues in one or both nuclease domains, you can create Cas9 proteins that nick (cut only one strand), bind without cutting, or perform entirely new chemistry. These variants have become the basis for the next generation of gene editing tools.

Cas9 Nickase (nCas9): Cutting Only One Strand

Mutating either catalytic residue (H840A to inactivate HNH, or D10A to inactivate RuvC) creates a Cas9 nickase — a protein that cuts only one strand of the DNA. A nick (single-strand break) is repaired with high fidelity by the cell, generally without introducing indels. By using two nickases with guide RNAs targeting opposite strands several base pairs apart, you can generate a double-strand break only at the intended site — dramatically reducing off-target effects, because two simultaneous nicks at off-target sites are statistically very unlikely.

Nickases are also the foundation of base editors and prime editors. Both use nCas9 (typically with D10A mutation, creating an HNH-active, RuvC-dead nickase) as their DNA-binding and nicking component. The nick is required for base editors to convert the deaminated base before the cell repairs it, and for prime editors to initiate reverse transcription of the edit template.

Dead Cas9 (dCas9): Binding Without Cutting

Mutating both catalytic residues (D10A and H840A) creates dCas9 (dead Cas9) — a protein that can find and bind any genomic target sequence with the same precision as wild-type Cas9, but cannot cut anything. dCas9 has become one of the most versatile tools in molecular biology because it functions as a programmable DNA-binding domain that can be fused to any effector protein you choose.

Section 10 — Putting It All Together: The Complete Cas9 Cycle

Let’s now walk through the complete CRISPR-Cas9 mechanism from start to finish as a unified sequence of events. Every step described above fits into this cycle:

References & Further Reading