Making a CRISPR knockout is now routine; making one you can defend is not. The nuclease is easy to deliver and cuts wherever a guide directs it, which means the experiment quickly reduces to three questions: did you cut the gene you intended, did the edit produce the allele you claim, and is the phenotype caused by the edit rather than by an off-target event or a clonal artefact. The design decisions and the validation strategy are what separate a useful knockout line from a story.
sgRNA design principles
The guide must be specific, efficient, and positioned to disrupt function. A few rules cover most cases:
- Position: target exons present in all annotated transcripts of the gene, ideally within the first two-thirds of the coding region. Cutting downstream of a functional domain may leave a truncated protein with residual activity.
- PAM and sequence: Streptococcus pyogenes Cas9 requires an NGG PAM, and the 20-nucleotide guide should have 40–70% GC content. Poly-T stretches can terminate transcription from a U6 promoter, and extreme GC content reduces cutting.
- Specificity: avoid guides with a second near-perfect match in the genome, particularly in a coding region. Seed-region mismatches (roughly the 8–12 nucleotides proximal to the PAM) matter most, but accumulating evidence shows that distal mismatches and even bulges can be tolerated at some loci.
- Functional constraints: avoid rare SNPs in your cell line, since a mismatch to the reference can prevent cutting; check the target sequence in the specific line you will edit, not just the reference genome.
Doench's Rule Set 2 and similar models predict on-target activity and are a reasonable starting filter, but the practical rule is to test two to four independent guides per gene and proceed with the ones that actually cut in your cells. Prediction accuracies are modest, and a guide that scores well in silico can fail at the bench.
Delivery: plasmid, mRNA, or RNP
How the components enter the cell determines the kinetics, the toxicity, and the persistence of editing.
- All-in-one plasmid expressing Cas9 and the guide: cheap, simple for easy-to-transfect lines, but slow to express, persistent, and prone to off-target activity because Cas9 remains active for days. Integration of the plasmid is a frequent and often unnoticed outcome, especially with selection.
- Cas9 mRNA plus guide RNA: faster expression and transient activity, reduced off-target risk, and no vector integration. Sensitive to RNase and requires careful handling; transfection efficiency can be lower than plasmid.
- Ribonucleoprotein (RNP): pre-assembled Cas9 protein with a synthetic guide RNA, delivered by electroporation or lipofection. RNP gives the fastest editing, the lowest off-target burden (the complex is degraded within hours), the lowest immunogenicity, and no DNA integration. It is the preferred format for primary cells and for any experiment where off-target editing is a concern, at the cost of higher reagent expense and a protocol requiring electroporation.
For hard-to-transfect cells, viral delivery (lentivirus, AAV) is an option, with the caveat that AAV has a small cargo limit and that lentiviral integration adds its own randomness. For most knockout projects, start with RNP; fall back to plasmid for cell lines where electroporation is impractical.
Detecting indels: T7E1, Surveyor, and NGS
Cas9 cleavage is repaired by non-homologous end joining (NHEJ), which creates small insertions or deletions (indels) at the cut site. Detection methods vary in resolution and sensitivity.
- Mismatch cleavage assays (T7 endonuclease I, Surveyor/CEL-I): PCR-amplify the region, denature and reanneal to form heteroduplexes between wild-type and mutant strands, then digest with an enzyme that cleaves mismatched duplexes. The resulting fragments are resolved by gel. This estimates the fraction of edited alleles and is fast and inexpensive, but it is semi-quantitative, misses edits that do not form a cleavable mismatch, and cannot resolve the exact sequence.
- Sanger sequencing of the pooled PCR product: an indel near the cut site produces a mixed trace that can be decomposed to infer the alleles present. It is the standard first look after single-clone isolation, but overlapping traces from a heterogeneous population are hard to interpret when more than two alleles are present.
- Next-generation sequencing of amplicons: the definitive readout. Amplicon sequencing reports the exact allele frequencies, distinguishes in-frame from frameshift indels, detects larger deletions, and quantifies editing efficiency. For knockouts, the crucial question is whether the indels produce frameshifts in the coding region and whether the mutant alleles are homozygous or compound heterozygous.
Whatever the readout, the primers must amplify the region and the assay must include a wild-type control and an unedited control. A common mistake is to declare a clone edited solely on the basis of a band shift, without confirming the sequence and the resulting protein.
Single-clone isolation and what cloning hides
A pool of edited cells is a mosaic, so functional experiments usually require clonal lines. Limiting dilution or single-cell sorting into 96-well plates isolates clones, which are then expanded and genotyped. Cloning introduces its own problems.
- Allelic complexity: clones may be heterozygous, homozygous, compound heterozygous for different indels, or carry multiple different edits. Genotype every clone and report the exact alleles.
- Compensatory adaptation: a clone derived from a single cell has undergone selection during expansion, and any phenotype of interest may reflect adaptation to the knockout rather than the knockout itself. Comparing several independent clones is the only way to distinguish.
- Off-target edits: cloning does not remove off-targets; it fixes them in the line. Sequencing candidate off-target sites in each clone, or running an unbiased method, is the way to address this. If all functional clones share the same off-target event and it is absent from unedited controls, it is a candidate confounder.
- Aneuploidy and drift: single-cell bottlenecking selects for clones that grow best, which may not be representative. Monitor karyotype and passage number, and compare against a wild-type clone from the same starting population that has undergone the same single-cell bottleneck.
Validating a knockout: the protein is the point
Genotype is a means; the question is whether the protein is gone. The validation ladder, in increasing strength:
- Amplicon sequencing confirming biallelic frameshift indels in the coding region.
- Loss of the protein by western blot with an antibody validated against a known positive sample, and confirmation that no truncated product is detected. Antibody specificity matters here: without a positive control, an absent band proves only that the antibody did not detect anything.
- Loss of function in an activity assay or phenotype, and — the strongest evidence — restoration by expression of a wild-type cDNA that is resistant to the guide (rescue).
- For genes with no functional assay, transcript-level evidence (loss of the mRNA or an altered sized product) plus sequencing may have to suffice, stated as such.
The rescue experiment deserves emphasis. Re-expressing the wild-type protein in the knockout background and restoring the phenotype demonstrates that the edit, and not a second-site mutation or clonal adaptation, caused the effect. Where the protein is essential, an inducible rescue or a hypomorphic allele may be the only viable approach. Reporting the guide sequences, the delivery format, the exact alleles, the clone identity, and the validation evidence is what makes a knockout publishable and reusable.
References
- Jinek M, Chylinski K, Fonfara I, et al. A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity. Science. 2012;337(6096):816–821.
- Cong L, Ran FA, Cox D, et al. Multiplex genome engineering using CRISPR/Cas systems. Science. 2013;339(6121):819–823.
- Mali P, Yang L, Esvelt KM, et al. RNA-guided human genome engineering via Cas9. Science. 2013;339(6121):823–826.
- Jinek M, East A, Cheng A, et al. RNA-programmed genome editing in human cells. eLife. 2013;2:e00471.
- Ran FA, Hsu PD, Wright J, et al. Genome engineering using the CRISPR-Cas9 system. Nat Protoc. 2013;8(11):2281–2308.
- Doench JG, Fusi N, Sullender M, et al. Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat Biotechnol. 2016;34(2):184–191.
- Kim S, Kim D, Cho SW, et al. Highly efficient RNA-guided genome editing in human cells via delivery of purified Cas9 ribonucleoproteins. Genome Res. 2014;24(6):1012–1019.
- Lin S, Staahl BT, Alla RK, Doudna JA. Enhanced homology-directed human genome engineering by controlled timing of CRISPR/Cas9 delivery. eLife. 2014;3:e04766.
- Vouillot L, Thélie A, Pollet N. Comparison of T7E1 and surveyor mismatch cleavage assays to detect mutations triggered by engineered nucleases. G3 (Bethesda). 2015;5(3):407–415.
- Brinkman EK, Chen T, Amendola M, van Steensel B. Easy quantitative assessment of genome editing by sequence trace decomposition. Nucleic Acids Res. 2014;42(22):e168.