CRISPR-Cas9 gene editing follows a sequence of carefully orchestrated steps: researchers design a short guide RNA that matches a specific DNA target, package it alongside the Cas9 protein, deliver both into living cells, and then let the cell’s own repair machinery finish the job. The whole process borrowed its logic from an ancient bacterial immune system, and understanding each stage reveals why the technology is both remarkably powerful and surprisingly tricky to get right.
Where the Idea Came From
Bacteria have been editing DNA for billions of years. When a virus infects a bacterium and the bacterium survives, it can capture a small piece of the invader’s DNA and store it in a region of its own genome called a CRISPR locus. That stored snippet becomes a molecular mug shot. If the same virus attacks again, the bacterium transcribes the snippet into a short RNA, pairs it with a cutting enzyme called Cas9, and the complex hunts down and destroys any matching viral DNA it finds.
This capture-and-destroy cycle is what researchers adapted for laboratory use. In nature, the Cas9 protein from Streptococcus pyogenes uses a two-part RNA guide: a CRISPR RNA (crRNA) that matches the target and a trans-activating RNA (tracrRNA) that helps the complex assemble. In 2012, researchers demonstrated that these two RNAs could be fused into a single “guide RNA” and paired with Cas9 to cut virtually any DNA sequence of interest.1PubMed Central. A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity That insight transformed an obscure bacterial defense mechanism into the most versatile gene-editing tool ever developed.
Step One: Choosing the Target
Every CRISPR experiment starts at a computer screen. You need to identify the exact stretch of DNA you want to edit, then design a guide RNA whose sequence matches that stretch. The guide is typically about 20 nucleotides long, and its sequence determines where Cas9 will land on the genome.
There is one non-negotiable constraint: the target site must sit right next to a short DNA motif called a PAM, or protospacer adjacent motif. For the most commonly used Cas9 (from S. pyogenes), the PAM is the three-letter sequence NGG. Without it, Cas9 will not bind or cut, no matter how perfectly the guide RNA matches the target.2PubMed Central. Structural basis of PAM-dependent target DNA recognition by the Cas9 endonuclease The PAM acts as a kind of molecular handshake: Cas9 first recognizes the PAM, and only then does it unwind the double helix and check whether the guide RNA matches the adjacent sequence.3PubMed Central. PAM identification by CRISPR-Cas effector complexes: diversified mechanisms and structures
Choosing a good guide RNA is not just about matching your target. It also means avoiding sequences that closely resemble other parts of the genome, because Cas9 can tolerate a few mismatches and cut in the wrong place. Computational tools now use machine learning to predict both how effectively a guide will cut its intended target and how likely it is to cause off-target damage elsewhere.4PubMed Central. DeepCRISPR: optimized CRISPR guide RNA design by deep learning Some platforms score individual guide-target pairings and aggregate them into an overall safety rating.5PubMed Central. Prediction of off-target activities for the end-to-end design of CRISPR guide RNAs In practice, most researchers design several candidate guides, test them experimentally, and pick the best performer.
Step Two: Assembling the Components
Once you have your guide RNA sequence, you need to produce the actual molecules and decide how to package them. There are three main formats for delivering CRISPR into cells, and the choice has real consequences for editing efficiency and cell survival.
- Plasmid DNA: A circular DNA molecule encoding both the Cas9 protein and the guide RNA is introduced into cells. The cell’s own machinery reads the DNA and builds the Cas9 protein and guide RNA internally. This is straightforward to produce but means Cas9 hangs around in the cell for a long time, which increases the window for off-target cuts.
- Messenger RNA: Instead of DNA, you deliver an mRNA transcript for Cas9 along with a separate guide RNA. The cell translates the mRNA into Cas9 protein, but because mRNA degrades relatively quickly, Cas9 production is temporary.
- Ribonucleoprotein (RNP): You pre-assemble the Cas9 protein and guide RNA in a test tube and deliver the finished complex directly. This is the fastest-acting format because the molecular scissors arrive ready to cut. RNPs also degrade within hours, which limits off-target exposure.
Head-to-head comparisons in difficult-to-edit cell types like mesenchymal stem cells show that RNP delivery consistently outperforms plasmid delivery, producing higher editing rates while keeping cells healthier.6PubMed Central. Highly efficient genome editing via CRISPR-Cas9 ribonucleoprotein (RNP) delivery in mesenchymal stem cells Similar results appear in blood-cell precursors, where RNP electroporation causes less cell death than plasmid electroporation.7PubMed Central. CRISPR/Cas9 ribonucleoprotein (RNP) complex enables higher viability of transfected cells in genome editing of acute myeloid cells
Step Three: Getting CRISPR Into the Cell
Having the right molecules assembled is only half the battle. You still need to get them inside living cells, and the delivery method depends on whether you are working in a dish or inside a living organism.
For cells grown in the lab, the most common approach is electroporation: brief electrical pulses open temporary pores in cell membranes, allowing the CRISPR components to slip through. Viral vectors, particularly adeno-associated viruses (AAVs), are another option and remain the go-to choice for many in vivo applications because they can home in on specific tissues. The trade-off is that AAVs have limited cargo capacity, which makes it difficult to fit the relatively large Cas9 gene inside them.
Lipid nanoparticles (LNPs) have emerged as a non-viral alternative, particularly for delivering CRISPR to the liver. These are the same type of fatty bubbles used in some mRNA vaccines, and they can encapsulate either mRNA or RNP payloads.8PubMed Central. Lipid nanoparticles: The game-changer in CRISPR-Cas9 genome editing For therapeutic applications outside the liver, though, targeting LNPs to the right tissue remains a significant engineering challenge.
Step Four: Finding and Cutting the DNA
Once inside the cell’s nucleus, the Cas9-guide RNA complex begins scanning the genome. Cas9 does not read every base pair sequentially. Instead, it slides along DNA and pauses whenever it encounters a PAM sequence. At each PAM, it pries open the double helix just enough to test whether the adjacent DNA matches the guide RNA.
If the sequences match, the guide RNA and target DNA form a structure called an R-loop: the guide RNA winds into the DNA, displacing one strand and base-pairing with the other. Structural studies show that Cas9 bends the DNA helix by about 30 degrees during this process, creating the physical distortion needed for the R-loop to form stably.9PubMed Central. Structures of a CRISPR-Cas9 R-loop complex primed for DNA cleavage The displaced strand is positioned near one of Cas9’s two cutting domains (called RuvC), while the strand paired with the guide RNA sits next to the other cutting domain (called HNH).
Cas9 does not cut immediately. A conformational checkpoint ensures that the R-loop is fully formed before the enzyme commits to cutting. The HNH domain must rotate roughly 140 degrees to reach its cutting position, and this rotation only happens when the match between guide and target is sufficiently complete.10Nature. R-loop formation and conformational activation mechanisms of Cas9 Once both domains are in position, each one snips its respective strand, creating a clean double-strand break at a fixed position three base pairs upstream of the PAM.
Step Five: Letting the Cell Repair the Break
A double-strand break in DNA is an emergency for the cell, and it triggers one of two major repair pathways. Which pathway kicks in determines the outcome of your edit.
The faster and more common route is called non-homologous end joining (NHEJ). The cell essentially glues the broken ends back together, but the repair is imprecise: small insertions or deletions (called indels) appear at the cut site. If the break falls within a gene, these indels often scramble the reading frame and knock the gene out. This is exactly what you want when the goal is to disable a gene. When two guide RNAs are used to make adjacent breaks, about half of the end-joining events rejoin the DNA accurately, but the rest introduce small errors, including characteristic single-base insertions.11PubMed Central. Harnessing accurate non-homologous end joining for efficient precise deletion in CRISPR/Cas9-mediated genome editing
The second route, homology-directed repair (HDR), is what researchers use when they want to write in a specific change rather than just break something. HDR requires a donor DNA template carrying the desired sequence, flanked by regions that match the DNA around the cut site. If the cell uses this template during repair, the new sequence is incorporated precisely.12PubMed Central. CRISPR-Cas9-mediated homology-directed repair for precise gene editing The problem is that mammalian cells strongly prefer NHEJ over HDR. Researchers have tried several tricks to tip the balance, including physically tethering the donor template to the Cas9 protein itself, which in one study boosted precise editing rates by up to 24-fold by concentrating the template near the break.13eLife. Covalent linkage of the DNA repair template to the CRISPR-Cas9 nuclease enhances homology-directed repair
Step Six: Checking Whether It Worked
After editing, you need to confirm that the intended change actually happened and screen for unintended damage. Several validation methods exist, and researchers often use more than one because each has blind spots.
The simplest and cheapest screening tool is a mismatch-detection assay using an enzyme called T7 Endonuclease I (T7E1). You amplify the edited region by PCR, let the strands re-anneal, and the enzyme chops any mismatched duplexes. It gives a quick readout of whether editing occurred but underestimates total editing rates compared to sequencing-based methods. When researchers compared T7E1 results with next-generation sequencing (NGS) at 19 genomic sites, NGS consistently revealed higher editing frequencies and detected a broader range of guide RNA activity.14The CRISPR Journal. Evaluation of Homology-Independent CRISPR-Cas9 Off-Target Assessment Methods Software tools like TIDE and ICE analyze standard Sanger sequencing traces and can identify and quantify individual indels, catching all major editing events above about 5% frequency at a fraction of the cost of NGS.15PubMed Central. Comparative Analysis of Methods for Assessing On-Target Gene Editing Efficiencies
Checking for off-target cuts is equally important. A method called GUIDE-seq works by flooding cells with short DNA tags that get captured at any double-strand break in the genome, not just the intended one. Sequencing around those tags reveals where Cas9 actually cut. When applied across 13 different guide RNAs in human cells, GUIDE-seq revealed wide variability in off-target activity and uncovered off-target sites that computational prediction alone had missed.16PubMed Central. GUIDE-seq enables genome-wide profiling of off-target cleavage by CRISPR-Cas nucleases
When Editing Goes Wrong at the Right Place
Off-target cuts get most of the safety attention, but on-target damage can be just as concerning. Long-read sequencing studies have revealed that a single Cas9 cut sometimes triggers deletions spanning thousands of base pairs around the intended site, far larger than the small indels researchers typically look for with standard short-read methods. Complex rearrangements, including inversions and even the joining of distant chromosomal segments, have been documented at intended cut sites in mouse embryonic stem cells and human cell lines.17PubMed Central. Repair of double-strand breaks induced by CRISPR-Cas9 leads to large deletions and complex rearrangements
In dividing cells, the picture gets worse. A Cas9-induced break can cause the chromosome to fragment during cell division, forming structures called micronuclei and chromosome bridges. In one study, this led to a scrambling pattern called chromothripsis on the targeted chromosome arm in the majority of cell lines examined.18bioRxiv. Chromothripsis as an on-target consequence of CRISPR-Cas9 genome editing These findings have pushed the field to develop editing strategies that avoid double-strand breaks entirely.
Editing Without Cutting
The risks of double-strand breaks motivated the development of base editors and prime editors, which modify DNA without fully severing both strands.
Base editors use a modified Cas9 that has been stripped of one or both of its cutting abilities and fused to a chemical enzyme that directly converts one DNA letter into another. Cytosine base editors (CBEs) convert C to T, and adenine base editors (ABEs) convert A to G.19PubMed Central. CRISPR-Cas9 DNA Base-Editing and Prime-Editing The deaminase enzymes that perform these conversions catalyze precise chemical changes at a single position, and because the DNA backbone stays intact, the large deletions and rearrangements associated with standard Cas9 are largely avoided.20PubMed Central. Nucleoside deaminases: the key players in base editing toolkit The limitation is that base editors can only make the four types of single-letter transitions they were designed for and cannot handle insertions, deletions, or the full range of possible substitutions.
Prime editors fill that gap. A prime editor pairs a Cas9 nickase (which cuts only one DNA strand) with a reverse transcriptase enzyme and uses an extended guide RNA called a pegRNA. The pegRNA contains both the targeting sequence and a template for the desired edit. After the nickase cuts one strand, the reverse transcriptase uses the pegRNA template to write the new sequence directly into the genome.21PubMed Central. Structural basis for pegRNA-guided reverse transcription by a prime editor Prime editing can install any small insertion, deletion, or substitution without a double-strand break and without a separate donor template, which makes it the most flexible precision-editing tool currently available.
Engineering Safer Scissors
In parallel with break-free strategies, researchers have re-engineered the Cas9 protein itself to cut more precisely. The wild-type enzyme tolerates several mismatches between the guide RNA and the target, which is why off-target cuts happen in the first place. High-fidelity variants like SpCas9-HF1 carry mutations that weaken Cas9’s grip on non-specific DNA contacts. The result is a protein that still cuts on-target with comparable efficiency for over 85% of guide RNAs tested but rarely cuts off-target sites.22PubMed Central. High-fidelity CRISPR-Cas9 nucleases with no detectable genome-wide off-target effects
A smaller Cas9 from Staphylococcus aureus (SaCas9) has also been engineered for higher fidelity. A single amino-acid change in its REC3 domain, which contacts the guide RNA-DNA hybrid, produced a variant that could distinguish single-base mismatches and reduced off-target activity by anywhere from about 2- to 93-fold across different sites.23PLOS Biology. High-fidelity SaCas9 identified by directional screening in human cells SaCas9’s smaller size also makes it easier to fit inside AAV vectors for in vivo delivery, which gives it a practical edge for therapeutic applications.
The Immune System Problem
One obstacle that rarely comes up in laboratory experiments but looms large for clinical use is the human immune system. The most widely used Cas9 comes from S. pyogenes, a bacterium that causes strep throat and other common infections. Most adults have been exposed to it at some point in their lives, and research has shown that a large fraction of the adult population already carries T cells that recognize and attack the SpCas9 protein.24Nature Medicine. High prevalence of Streptococcus pyogenes Cas9-reactive T cells within the adult human population This pre-existing immunity could cause the body to destroy Cas9-edited cells before the therapeutic benefit takes hold, or worse, trigger a dangerous inflammatory reaction.
This is one reason the field is diversifying away from SpCas9. Cas9 proteins from bacteria that rarely infect humans, engineered variants with altered surface features, and transient delivery formats that clear Cas9 from the body quickly are all being explored to sidestep immune recognition. For ex vivo therapies, where cells are edited outside the body and then returned to the patient, the immune problem is less pressing because Cas9 does not need to persist inside the body.
Beyond DNA Cutting
The CRISPR platform has expanded well beyond its original cut-and-repair function. A catalytically dead version of Cas9, called dCas9, can no longer cut DNA but still binds to a target guided by its RNA. Fusing dCas9 to different functional proteins turns it into a Swiss army knife for gene regulation. Attach a transcriptional activator, and dCas9 turns a gene on. Attach a repressor, and it turns a gene off. Attach an epigenetic modifier, and it can add or remove chemical marks on DNA or its associated proteins without altering the genetic sequence at all.25PubMed Central. Genetic and epigenetic control of gene expression by CRISPR-Cas systems These “epigenome editing” tools are particularly appealing because the changes they make are potentially reversible, unlike permanent DNA edits.
CRISPR has also moved beyond DNA targets altogether. The Cas13 family of enzymes targets RNA rather than DNA, enabling researchers to knock down gene expression at the transcript level without touching the genome. Because no permanent change is made to the cell’s DNA, Cas13-based approaches avoid the risk of heritable off-target mutations entirely.26PubMed Central. CRISPR-Cas13: Pioneering RNA Editing for Nucleic Acid Therapeutics RNA-targeting systems are being explored for applications where temporary gene silencing is the goal, such as reducing the expression of a disease-driving protein during a flare-up without permanently disabling the gene that makes it.