How CRISPR-Cas9 editing works in research

CRISPR-Cas9, as most research benches mean it, is an RNA-guided nuclease. A guide RNA pairs with a DNA target next to a short motif the protein will accept, and Cas9 cuts both strands there. The cell then repairs the break. The edit you actually get is the product of that repair, not a photocopy of the guide drawing. This page explains that mechanism well enough to plan a research experiment and to see why verification is part of the design. How to read the result is in checking whether a genome edit worked.

It is a research explainer. It is not a therapeutic protocol, not a kit insert, and not permission to edit a human germline or to bypass institutional review. Reagent classes are discussed through the molecular biology catalogue and the quote request.

What the protein and the RNA are doing

The widely used Cas9 from Streptococcus pyogenes recognises a PAM, commonly written NGG, on the target DNA. The guide RNA is complementary to the stretch of target just upstream of that PAM. In many experiments the guide is a single RNA that fuses the targeting sequence to the scaffold the protein binds. A two-RNA version is the same idea in an older format. Without the PAM this enzyme uses, a matching sequence is not a substrate. A guide designed for SpCas9 is not a plan for a different Cas protein.

When the RNA-protein complex binds and the match is good enough, Cas9 cuts both strands, typically leaving a blunt break a few bases upstream of the PAM. The cut is the reagent. A knockout or a knock-in is a repair outcome, and it has to be earned with data.

Design starts from a named genome assembly. Ensembl and GenBank are public places those sequences live. A guide drawn on an old assembly can land in the wrong exon or on a variant your cells do not have. Record the assembly identifier in the notebook next to the guide sequence. Predicted on-target scores are useful for ranking guides and useless as a promise that this guide will cut in this chromatin.

Off-target sites are similar sequences, especially those that keep a PAM and differ in positions the protein tolerates. Mismatches in the seed region, close to the PAM, are often more disruptive to cutting than mismatches at the far end of the guide. Chromatin, dose and how long the nuclease stays in the cell all move the real pattern. A design tool's list is a set of sites to consider, not a map of everywhere the enzyme can act. The Addgene CRISPR guide is a public overview of these research concepts. It is not a protocol to copy volumes from, and this page does not add those volumes.

What the cell does with the break

Without a repair template, many mammalian experiments repair the break by end joining. The cell sticks the ends back together and often adds or removes a few bases. Some scars shift the reading frame. An in-frame deletion of three bases can leave a protein that still works. A transfected guide is not, by itself, a knockout.

A mixture is the normal first result. One cell can carry two different scars. Clonal isolation is how you turn a mixture into a genotype you can name. Bulk DNA gives an average, and it is honest only if you call it an average.

Homology-directed repair can copy from a donor template that carries homology arms flanking the change you want. That path is linked to the cell-cycle stage and is often less frequent than end joining. The donor can be a plasmid or a single-stranded DNA, and the design of the arms is part of the experiment. Silent changes that block the guide from recutting the repaired allele are a common design habit, because an already-correct allele can otherwise be cut again and scarred. HDR is the wrong hope when you only needed a frameshift and end joining would have answered the question with less machinery.

Base editors and prime editors are different designs. If your experiment is a nuclease cut, analyse it as a nuclease cut.

Delivery is part of the dose

You can deliver Cas9 and the guide as DNA on a plasmid, as RNA, or as a preformed ribonucleoprotein. A plasmid can keep expressing nuclease, which gives off-target sites more time. A ribonucleoprotein acts and is degraded, which suits a pulse of cutting. Choose the class for the exposure you want. Transfection toxicity that kills the edited cells biases the genotype you recover.

Selection markers and fluorescent reporters help you enrich transfected cells. They do not prove the locus changed. A bright reporter with a wild-type target site is a common and disappointing gel. Plan the assay that reads the locus, not only the assay that reads the delivery marker.

Controls belong in the same transfection. A no-guide or non-targeting guide preparation shows what the delivery stress does alone. An untransfected culture shows the baseline alleles. A positive guide against a locus you have already edited successfully, when you have one, shows that this batch of protein or plasmid can cut in these cells. Without that, a wild-type read can mean a dead reagent or a real miss.

Off-target risk is a design choice

You cannot make the risk zero by wanting the guide to be specific. You can avoid obvious extra sites, keep the nuclease exposure limited, and decide which off-target checks the claim requires. A research tool in a cell line, used to suggest a phenotype, is not held to the same evidence standard as a strain that will be shared for years. The standard should be written down before the experiment, not negotiated after a surprising phenotype.

Methods that look for breaks without a predicted list are extra experiments, with their own detection floor. Naming them is not the same as having done them. For many projects the honest sentence is narrower: these predicted sites were sequenced, and genome-wide absence of off-target cuts was not tested.

Mosaicism matters in animals and in any culture that was not cloned. One well can contain edited and unedited cells. The verification page treats that as a measurement problem.

Safety, climate and the enquiry

Edited cells and animals remain biological agents. The guide and the donor do not set the biosafety level. The organism, the insert, and the local committee do. The WHO laboratory biosafety manual is a public reference for how laboratories frame that decision. This article does not assign a containment level, and it does not authorise work on human embryos or on pathogens outside the approval you already hold.

Guide RNA is RNA. In a hot laboratory it follows the same impatience as any RNA reagent: keep it cold when the format requires, and do not assume a tube that sat on a shared bench is the molecule you designed. Cell incubators that lose power mid-recovery change viability and can change which clones you recover. Record the excursion. Do not fold those plates into the main series without a note.

EVRINTH can take a sourcing question about nuclease, RNA and plasticware classes. State the organism, the delivery class, whether you need a donor, and how you will verify the locus. The CRISPR validation sequencing enquiry reference is a prompt for the sequencing part of that plan. It is an independent method reference. Ask whether a quotation is possible. Do not read it as a statement that an editing service is already running. The molecular biology pathway is the wider experimental context.