Start with the biological question, not the technique

CRISPR has become so routine in UK labs that it is easy to treat it as a single, off-the-shelf method. It is not. The tools you choose, the delivery route you pick and the validation you plan all flow from one decision: what exactly do you want the cells to do? A clean knockout, a precise point mutation, a tagged endogenous protein and a transcriptional reporter are four quite different projects, and each carries its own practical risks.

Before ordering anything, write down the edit in the most literal terms you can. If you intend to disrupt a gene, decide which exon to target and what the residual transcript might do — nonsense-mediated decay is not guaranteed, and residual protein fragments can retain function. If you intend to insert a sequence, decide the exact junction, whether you need a selection cassette, and how you will remove it later. Spending a fortnight on this saves months at the bench.

Choosing an editing strategy that fits your system

Most academic projects use one of a small number of approaches, and it helps to be honest about their limitations.

• Standard Cas9 nuclease knockout. Creates a double-strand break repaired by non-homologous end joining, producing frameshift indels. Fast, cheap and usually sufficient for loss-of-function work in immortalised lines.
• Nickase or paired-nicking strategies. Reduce unwanted larger rearrangements, useful when you are working with cells that are prone to genomic instability.
• Base editors. Convert C to T or A to G without a double-strand break, ideal for modelling single-nucleotide variants or installing premature stop codons. They are constrained by PAM placement and editing window, so design is less flexible.
• Prime editing. Handles small insertions, deletions and all twelve substitution types, but efficiency varies widely between loci and cell types. Treat it as a project in itself rather than a fallback.
• CRISPR interference and activation. Catalytically dead Cas9 fused to repressor or activator domains lets you modulate transcription without changing the genome — often the fastest route to a phenotype in primary or difficult-to-edit cells.

Cas12a is worth considering for AT-rich regions where suitable Cas9 protospacer adjacent motifs are scarce, and it processes its own guide arrays, which simplifies multiplexing.

Designing and pre-testing your guides

Spacer design is still where most projects quietly fail. Aim for a 20-nucleotide spacer with 40–60% GC content, avoid long homopolymer runs and strong secondary structure, and check that the target site is conserved across the transcripts you care about. Crucially, check your own cell line for single-nucleotide variants in the seed region: a common polymorphism can abolish editing on one allele and leave you puzzling over heterozygous results.

• Select three to five candidate guides per target rather than one, and test them in parallel.
• Rank candidates with more than one in silico scoring algorithm; agreement between tools is more informative than a single high score.
• Confirm activity cheaply before committing — an in vitro cleavage assay with recombinant Cas9 and a PCR amplicon takes an afternoon.
• Always run a non-targeting guide control through the same workflow, and keep an unedited parental population alongside.

Delivery: matching method to cell type

Delivery is the variable that most often dictates success. Lipid-based transfection of plasmid DNA is inexpensive and works well in HEK293 and other permissive lines, but it is inefficient in primary cells, and prolonged plasmid expression increases off-target activity and the risk of random integration.

Electroporation of pre-assembled ribonucleoprotein complexes is the current default for most demanding systems, including induced pluripotent stem cells, primary T cells and haematopoietic progenitors. The edit is transient, toxicity is lower than with plasmid, and you can re-electroporate if efficiency is poor. Lentiviral delivery gives stable Cas9 expression and works where nothing else does, but it requires containment level 2 facilities and careful titre control. For zygotes and early embryos, microinjection remains the standard, while adeno-associated vectors suit some primary and post-mitotic cells.

If you need homology-directed repair, remember that it competes with end joining. Use a single-stranded oligodeoxynucleotide donor for short edits, block the PAM or the guide site after successful editing to prevent re-cutting, and consider synchronising cells or adding small-molecule enhancers. Expect precise integration rates to be lower than indel rates, often substantially.

Validating the edit properly

An indel detected by a gel is not proof of a functional knockout. Genotyping should combine Sanger sequencing with indel-tracking analysis or, better, amplicon next-generation sequencing to quantify allele frequencies in the pool. For clones, sequence both alleles and check for larger deletions or loss of heterozygosity.

Then validate at the protein level by Western blot with an antibody you have verified, and ideally with a functional or phenotypic readout. Off-target assessment should start with in silico prediction, followed by targeted amplicon sequencing of the top-ranked sites; unbiased whole-genome sequencing is reserved for cases where the stakes justify the cost. Karyotype or check copy number in any clone you intend to bank, particularly after single-cell dilution, and confirm identity by short tandem repeat profiling.

Governance and approval in UK institutions

Regulatory paperwork is not an afterthought. Work with genetically modified organisms is governed by the Genetically Modified Organisms (Contained Use) Regulations 2014, administered by the Health and Safety Executive. Your local genetic modification safety committee will classify the activity, usually class 1 for disabled laboratory strains and class 2 where the modification could pose a hazard. Allow time for committee review, and note that notification to the HSE is required before class 2 work begins.

Human tissue and primary samples bring additional duties under the Human Tissue Act, with approval from a research ethics committee and, where relevant, the Human Tissue Authority. Research involving human embryos falls under the Human Fertilisation and Embryology Authority, and the 14-day limit is absolute. Animal work requires a Home Office project licence under the Animals (Scientific Procedures) Act 1986, plus personal and establishment licences. Standardised reporting, such as the ARRIVE guidelines, is expected by most funders.

Keep contemporaneous records of guides, plasmids, cell lines and passage numbers, and archive them for the retention period your institution specifies. Good bookkeeping is what allows the next person in the lab — or you, two years from now — to reproduce the work.