Controls for a CRISPR experiment

The well is green, the antibiotic has killed the untransfected cells, and the figure legend already says knockout. None of those observations is the edit. Controls for a CRISPR experiment are the wells and the records that keep delivery, stress, cutting, and phenotype from being the same sentence. The mechanism they are controlling is in how CRISPR-Cas9 editing works in research. The sequence standard they have to meet is in checking whether a genome edit worked.

What you are trying not to confuse

A CRISPR well differs from its parent in several ways at once. Nucleic acid or protein entered the cell. A scaffold RNA engaged a nuclease. Selection or sorting was applied. The culture grew for extra passages. Any of those can change a phenotype. The cut at the intended locus is only one candidate. Controls peel the candidates apart. They are not a moral extra. Without them, a later sequencing result has nothing calm to be compared with.

Place the controls in the same plate or the same day's transfection as the experimental guide. A control run last month on a different passage answers last month's question.

The parent, written down

Parental identity is the reference for every claim of difference. Record the line or strain name, where the vial came from, the passage or the colony stock, and at least one fact you can recheck: a known SNP, a documented karyotype note from your facility, or the antibiotic the parent is supposed to resist. Misidentified lines are common enough that a famous name on the flask is not a genotype. Ensembl can show you the reference the guide was drawn on. Your cells may not match it. If they do not, the PAM may be missing and a "failed edit" is a mismatch between flask and browser.

Keep an unedited aliquot. Once every tube has been through the nuclease, you cannot reconstruct the parent from memory.

Untransfected cells

An untransfected, or untransduced, sister culture goes through the same medium, the same time, and the same splits, without guide and without nuclease. It catches contamination, medium mistakes, and drift. If the phenotype appears here, it is not your spacer.

If selection is part of the protocol, the untransfected cells should die under that selection when the marker works, and they should live when you forget the drug. That pair tells you the drug is real. It still does not tell you the chromosome changed. A resistance gene on a plasmid can protect a cell that never cut the target.

A non-targeting guide

A non-targeting guide uses the same scaffold, the same nuclease, the same delivery, and the same selection as the experiment, with a spacer chosen not to match the genome under the enzyme's rules. Design tools such as CRISPOR make that comparison explicit. The well absorbs the stress of transfection and of having a nuclease in the cell. If the phenotype appears in this well at the same strength as with your guide, the spacer is not a sufficient explanation.

Non-targeting is not the same as a scrambled sequence you invented without searching. A casual scramble can land on a real site next to a compatible motif. Search it. Also accept the limit: absence of a perfect match is not absence of every partial match. If the non-targeting well behaves oddly, sequence the locus you care about anyway, in case the oddness is unrelated culture stress, and treat a gene claim from that batch as blocked until the control behaves.

A guide that is already known to cut

Include one guide that has cut a control locus in this cell type with this delivery class. The locus can be a safe, previously characterised site your laboratory already knows how to amplify. The point is the positive control for the method, analogous to a positive control in PCR controls and contamination control.

If the known guide produces no indel and your new guides also produce none, you do not have evidence against the new spacers. You have a delivery failure, a dead nuclease, a wrong PAM enzyme, or a detection assay that cannot see indels. Fix that chain. If the known guide cuts and the new one does not, then look at the spacer, the chromatin, and the match between the guide and this parent's sequence.

Detect the known cut with the same assay you trust for the experiment. A control you only believe because a paper once saw it, and that you did not remeasure, is a citation, not a control.

More than one guide

For a claim about a gene's phenotype, use at least two guides that cut different positions and, when you can, different strands or different exons that still test the same function. Sequence each to show the on-target allele. A phenotype that tracks both guides is harder to explain as one off-target site. A phenotype that tracks only one guide is a reason to stop and look at that spacer's other possible sites, or to drop the claim.

Two guides that both fail to cut are a locus problem or a design problem, not two independent confirmations of resistance to editing. Cutting is established per guide.

In a pooled screen the same idea appears as several spacers per gene, with non-targeting and essential-gene behaviour as population controls. That layout is compared in arrayed versus pooled screens.

Rescue as a concept

Rescue means you put back the function you think you removed, in a form the guides should not destroy, and you ask whether the phenotype returns toward the parent. A cDNA with silent changes at the cut site, or a cDNA expressed outside the genomic locus the guides target, is the usual shape of that idea. Design it so the guides' spacers do not simply cut the rescue construct. If they do, a failed rescue is uninterpretable.

Rescue supports a gene-level story. It does not by itself prove the alleles. You still sequence the locus. A rescue that fails does not prove an off-target either. The cDNA may be poorly expressed, the tag may block the protein, or the phenotype may need a particular isoform. Say which of those you checked.

Do not describe a fluorescent marker as a rescue. The marker was never the missing gene.

Branching when a control fails

Known cutter silent, experimental guides silent: treat it as a methods failure. Check the enzyme lot, the guide integrity, the delivery settings the manufacturer or your core specifies, and the detection PCR. Do not redesign twenty spacers first.

Known cutter positive, experimental guide silent at sequence: the spacer, the parent's local sequence, or chromatin is the branch. One new design is rational. Ten new designs without looking at the parent's PAM are not.

Phenotype in the non-targeting well: stop the gene claim for that batch. Look for contamination, dosage of nuclease, and whether the non-targeting spacer matches this genome better than you thought.

Phenotype in only one of two cutting guides: report the alleles, and do not write a gene-level sentence yet. Consider an off-target or a clone-specific passenger mutation. Rescue, if the construct is clean, is the next conceptual test, not a longer incubation.

Untransfected cells show the phenotype: the variable is outside the nuclease. Change the claim or find the contaminant.

Safety and what controls do not approve

Controls do not authorise the experiment. Biosafety and ethics decisions sit with the institution and outside any reagent quotation. A non-targeting guide is still a nucleic acid introduced into cells. A viral delivery of a control guide is still viral delivery. Research controls are not a diagnostic panel and not a clinical release test.

Shared freezers make parental identity a practical problem. A vial inherited from another building needs its name checked before it becomes the baseline for a knockout. In a humid lab, labels peel. Write the control identity on the lid and in the book. A power cut during a transfection is a reason to discard that plate's comparison, not a reason to compare it with next week's controls.

Ordering with the controls named

Tell the quote request which nuclease class, which guide format, and whether you need a non-targeting sequence and a known control locus as well as the experimental spacers. Oligonucleotides and enzymes are classes in the molecular biology catalogue. Confirmation sequencing, once the controls say a cut is plausible, can be discussed from the CRISPR validation sequencing enquiry reference, an independent method reference. Ask whether a quotation is possible. The molecular biology pathway is the research context for the assay. Name the controls in that note so the materials and the experiment stay the same shape.