Checking whether a genome edit worked

A genome edit worked only in a specified sense: this allele, in this cell or this clone, has this sequence. A transfected well is not that sentence. Neither is a single PCR band. This page is the order of checks that keeps those sentences apart. The mechanism of cutting and repair is in how CRISPR-Cas9 editing works in research. What follows is how you decide the result is real. It is a research explainer, not a regulated assay.

Primers, polymerases and sequencing consumables are catalogue classes in the molecular biology catalogue. The verification plan belongs in the quote request.

Start with the locus, and a PCR that can be wrong in an informative way

Design primers on a named reference, using a tool such as NCBI Primer-BLAST, so you know what else they might bind. Place them outside the window where a small scar or a modest deletion would sit. If both primers land inside a region the cell might delete, the deleted allele gives no product and the only band you see is the unedited allele that remains. You will call a heterozygote wild type, or call a failed PCR, depending on how unlucky the design is.

Run the unedited sister culture beside every edited sample. Run a no-template control. The rules are the same as in PCR controls and contamination control. A synthetic donor or a previous amplicon of the edited allele is a potent contaminant. If the no-template lane shows the product, you do not have evidence about the cells.

Look at the size, and then do not stop there. A large deletion can shift a band in a way an agarose gel will show. A one-base insertion will not. A smear or a doublet of PCR products can be a real mixture of alleles or a tired PCR. Clean the reaction and sequence it before you draw the deletion. Gel conditions that make those judgments fair are in agarose gel electrophoresis for DNA.

What a sequence is allowed to say

Sanger sequencing of a clean amplicon from a diploid clone should be readable through the cut if both alleles match. Peaks that split at the cut and stay split mean the alleles differ. That pattern is the result, not a failed run, provided the trace was clean before the split. Tools that decompose mixed traces can propose which indels are present and roughly how much of each. Treat the proposal as a lead. Confirm by separating the alleles: clone the amplicon into a plasmid and sequence several colonies, or sequence the amplicon deeply enough to count molecules.

Bulk cultures and mosaic tissues are mixtures by default. A single genotype call from a bulk Sanger trace overstates the data. Deep amplicon sequencing is the method class that counts those molecules. It has its own errors, including PCR duplicates and polymerase mistakes that look like rare indels. A frequency near the error floor is not an allele. The public overview of short-read sequencing from Illumina is background for that method class, not a pipeline for your samples.

Homology-directed repair needs a stricter read. You are asking whether the donor sequence is present at the junctions. End joining also destroys the wild-type cut site. A silent blocking change you designed into the donor should appear in a true repair and be absent from a random scar.

Record the reference assembly, as you would have done at design time on Ensembl. A variant that is already in the parental line is not an edit. Sequence the parent.

Knockout, knock-in and clone

Write the noun you mean. A sequenced frameshift on one allele is a heterozygous frameshift. It becomes a homozygous frameshift only when every allele in that genome shows it. Cancer lines and many cultured lines are not tidy diploids. If the parent has three copies, three alleles have to be accounted for. A western blot or a targeted protein assay can show that the product is missing, which a DNA trace cannot. Antibody limits still apply: a missing band means this antibody no longer sees the protein, under conditions where a control lysate still works.

A knock-in claim names the bases that should have changed and shows they are present at the locus, in the alleles you say they are in. Random integration of a donor plasmid elsewhere in the genome can create a PCR product if your primers are careless, and it can create a drug-resistant colony that never touched the target locus. Include a primer pair that cannot amplify the donor alone.

A well that began as one cell can still pick up a second clone. A trace that becomes mixed after having been clean is a contamination or a mixed culture until you reclone and sequence again.

Off-target checks without inflating them

A predicted-site panel is a PCR and a sequence at each site you chose to worry about, with the same no-template discipline. Absence of a scar at those sites is a real, limited result. It does not become genome-wide by adjectives.

Unbiased methods look for breaks or repair scars without depending only on a homology search. They have detection floors. A method that needs a certain allele fraction will miss a rare clone. If you did not run such a method, do not cite its existence as if it had cleared your guide. The Addgene CRISPR materials describe these research choices at a high level. They are not a substitute for the figure you actually have.

A phenotype that appears in a pool and disappears in every clean clone may be off-target, a passenger mutation, or a response to the transfection. Believe the clones, or show that putting the on-target allele back restores the effect.

Records, climate and safety

Keep the primer sequences, the reference coordinates, the raw traces and the gel image with the control lanes. A cropped band is not a record. If a power cut or a network drop separates the sequencer from the notebook, the run still needs a file name that maps to the clone. Reconstructing genotypes from memory is how heterozygous lines become "knockouts" in a slide.

Edited material stays under the biosafety and ethics approval that covered the editing. Verification PCRs on that DNA are not automatically a lower category if the specimen itself is hazardous. This article does not change that approval and does not provide a diagnostic interpretation of anyone's genome.

Heat and contamination interact. A shared PCR room in a warm building, with donors and amplicons in the same freezer, is how an edited band appears in the unedited control. Physical separation, described in the PCR controls note, is part of edit validation. If the unedited control ever shows the donor sequence, stop and treat the assays as contaminated.

What to ask when sequencing is the bottleneck

EVRINTH can take a sourcing question. State the organism, the locus coordinates, whether the sample is a clone or a bulk population, the amplicon length, and whether you need Sanger traces or a counted deep read. The CRISPR validation sequencing enquiry reference is a prompt for that specification. It is an independent method reference. Ask whether a quotation is possible. Do not send only a gene name and a wish for "confirmation". The molecular biology pathway is the surrounding experimental context.