CRISPR interference and activation overview

Turning a promoter down with a dead Cas9 is not the same experiment as cutting the gene. Turning it up is not the same experiment as inserting a cDNA. CRISPR interference and CRISPR activation change transcription while the machinery is present, and they leave the locus sequence alone unless something has gone wrong. This page is a troubleshooting guide for those two designs. The nuclease experiment people confuse them with is in how CRISPR-Cas9 editing works in research. What a real protein loss looks like, when that is actually the claim, is in indels, frameshifts and protein loss.

Expression constructs and guide RNA are reagent classes in the molecular biology catalogue. The fusion you intend should be named on the quote request.

Dead Cas9 is a binder

Catalytically dead SpCas9 carries mutations that silence both nuclease domains, the familiar pair being D10A and H840A together. The protein still uses a guide and a PAM to sit on DNA. It does not, in the design you want, cut either strand. A plasmid that has only one of those mutations is a nickase. A plasmid that has neither is a nuclease. Both will produce a DNA-damage experiment and a transcriptional story that does not add up. Read the sequence of the fusion you received. Do not trust a nickname on the tube.

Because the protein must remain bound or keep recruiting, the effect tracks expression of the fusion and the guide. Plasmid loss, silencing of a viral insert, or uneven expression across a pool all become changes in the phenotype. That is the opposite of a frameshift, which stays after the nuclease is gone.

Interference: KRAB at the start site

The widely used repressor is dCas9 fused to a KRAB domain. KRAB recruits co-repressors and can lay down local heterochromatin in addition to the steric block of the Cas protein sitting on the DNA. dCas9 without KRAB can hinder elongation if it is tiled in a gene body, and it is usually a weaker way to silence a promoter than the KRAB fusion. When people say CRISPRi in current mammalian work, they usually mean dCas9-KRAB.

Guide position is the first troubleshooting variable. Strong repression is typically sought in a window around the transcription start site. A practical planning window is from roughly 50 bases upstream of the start to about 300 bases downstream, and many of the strongest guides land just downstream of the start. That window is a planning range from the method literature, not a promise for every promoter. Guides in the middle of a long intron often do little. Guides on the wrong isoform's start site do nothing to the transcript you are measuring.

Repression is partial. A useful guide is one you measured, by RNA or by protein, not one a score merely ranked. Some genes fall ten-fold. Some fall by a third. Some do not move. Write the magnitude. A gene that still produces a third of its protein is not a null, and a phenotype that required a null may not appear. Compare that honestly with a cutting knockout rather than renaming the knockdown.

Bidirectional promoters and closely spaced genes are the neighbour problem. KRAB can repress more than the gene you signed up for. If the phenotype could come from the gene next door, measure the neighbour. Isoforms with different start sites need different guides. Ensembl is where those starts are annotated, and the annotation can be wrong in your cell type. A quick look at which start is used, even by existing RNA data, saves a month of flat qPCR.

Activation: VP64, VPR and SAM

First-generation activation fused VP64, a cluster of VP16 activation domains, to dCas9. It works modestly at many genes and poorly at others. VPR adds further activation domains, p65 and Rta, in one fusion, and is stronger in many published comparisons. SAM keeps dCas9-VP64 and changes the guide: aptamers in the scaffold recruit a second protein, MCP fused to p65 and HSF1. SAM is three components, not one plasmid. Forgetting the helper protein is a common reason SAM "does nothing".

Activation guides are usually placed upstream of the transcription start, often within a few hundred bases, not on top of the start the way a repressor guide might be. Again the range is a planning window. Tiling several guides in that window is a published tactic when one guide is weak. A single guide that you have shown increases the transcript is enough for a focused experiment. A screen will have its own library rules.

Activation still depends on the endogenous locus. A cDNA overexpression plasmid does not. If the promoter is deeply silent, recruitment may not open it. If the gene is already highly expressed, a further fold change may be small. CRISPRa does not install a variant allele and does not add a tag. Those remain knock-in problems.

SunTag and other recruitment scaffolds exist as further classes. Do not mix their component lists. A SAM aptamer guide with a SunTag antibody system is an accident, not a hybrid method.

When the expression measurement is flat

Measure the editor before you blame the gene. An antibody or a fluorescent fusion should show dCas9 in the cells you are scoring. A guide-expression cassette that was lost during cloning will give you a protein with nothing to bind. Confirm the spacer against the genome build in GenBank or Ensembl, on the strand convention the library uses.

Then check the start site. A guide designed to a start that this cell type does not use will sit in dead sequence. Move the guide to the start the RNA actually comes from. Then check dose and time. Transient transfection can be too brief or too uneven to see a stable knockdown. Stable expression, often lentiviral in screening lines, has the opposite risk: the insert silences over passages. Re-measure RNA at the passage of the assay, not only at the passage where you made the line.

The assay has its own floor. Small fold changes need the discipline in RT-qPCR for relative expression, inside the larger path of from cells to a gene expression result. A noisy primer pair will invent both success and failure. Protein is the right second measurement when the claim is about the protein. RNA movement without protein movement is an assay result, not a CRISPRi failure by itself, and it should be reported as such.

If the cells die only with one guide, consider an off-target cut from a nuclease contaminant, or real on-target essentiality of a strong knockdown. Sequence the locus if you have any doubt the protein is dead. A cut changes the interpretation completely. The delivery choice for keeping expression stable is discussed as a class in delivering editing reagents to cells. Lasting viral expression is often what CRISPRi and CRISPRa want, and what a one-time nuclease edit should avoid.

Neighbours, isoforms and a nuclease used by mistake

A phenotype with no change in the target RNA is not CRISPRi until you have shown the target RNA should have changed. Measure it. A phenotype with a change in the neighbour and not the target is a neighbour effect until proven otherwise. A phenotype that persists after you have removed the guide is no longer explained by CRISPRi. Look for a cut, a passenger mutation, or a selected subclone.

Essential genes can be studied with CRISPRi precisely because the knockdown can be partial and, in principle, reversible. If you then select for cells that grow despite the guide, you may be selecting for cells that silenced the editor. Growth is not evidence the gene was dispensable. Check that dCas9 is still expressed in the cells you kept.

Do not combine a cutting guide library with a dCas9 line and call the dead wells knockouts. And do not combine a dCas9 guide library with nuclease Cas9 and call the result transcriptional. The protein defines the experiment. The guide sequence does not.

Overexpression is an institutional risk question

CRISPRa can drive a gene hard, including genes that affect growth, signalling or viral susceptibility in a research model. That is a reason to name the target in the institutional risk review, not a reason this page can rank targets as safe. CRISPRi of a hazardous function does not make the cell harmless. Approval covers the line, the delivery and the genetic modification. The WHO laboratory biosafety manual is general laboratory background. This article is not a permit, not a diagnostic method, and not a procedure for embryo or germline editing.

Viral delivery of dCas9 is still viral delivery. Follow the containment assigned to the vector. Dead nuclease does not mean dead biosafety review.

Keeping the construct present

The practical failure mode in a long experiment is silent loss of expression. Antibiotic selection on the vector can slow that loss and can also hide it if the resistance gene and the fusion are not equally stable. Re-check the fusion at the passage you assay. In a warm laboratory, incubator drift and a power cut stress cells and change transcript baselines, so a fold change measured the day after an outage is not comparable to the week before. Repeat the RNA measurement when the culture has recovered. Humidity and peeled labels on guide plasmids are how a repression guide and an activation guide trade places. Sequence the spacer from the plasmid in the cells if a line suddenly does the opposite of what you designed.

Store expression plasmids and RNA as their own conditions require. A KRAB fusion and a nuclease Cas9 prep are different tubes even if both say Cas9 on a casual label.

What a regulation enquiry needs

State CRISPRi or CRISPRa, the fusion class (KRAB, VP64, VPR, or SAM with its helper), the cell type, and whether expression must be stable or transient. Include the transcription start you are targeting and how RNA or protein will be measured. Do not describe the job as a knockout. Sequencing is the wrong default readout unless you are checking that the locus was not cut. If you do need that check, the CRISPR validation sequencing enquiry reference is a way to discuss it. The page is an enquiry reference and does not mean a service is operated. Expression reagents sit in the context of the molecular biology pathway. Ask whether a quotation is possible. A request that says CRISPR knockout will be packed for a different experiment.