Life Finds a Way - Genotype-Phenotype Discord in genome editing

Why relying solely on phenotypic gene editing confirmation is risky.

One of the hallmarks of every living system is that it finds a way to overcome deficiencies in order to survive and thrive. When genome edits are performed, these often introduce new deficiencies that the cells will try to overcome in order to thrive. For this reason, investigations that rely solely on phenotypic characterization for functional validation of a genome edit are risky.

The accuracy of gene editing is critical to the validity of scientific insights

Genome editing is rapidly demonstrating its power for understanding and manipulating the sequences of organisms. It has become relatively straightforward to make precise changes such as insertions, deletions, and individual nucleotide changes to a sequence of DNA and characterize the resulting phenotype.

As a crude indicator, over 5500 publications are listed when a search is performed using the terms “phenotype” and “CRISPR” in the PubMed database; this only scratches the surface since these two terms imperfectly capture the breadth of research in this field. While the correlation between the designed genomic alteration and expected phenotype is often good, buried within these examples are instances where life has indeed found a way to overcome the researcher’s intent.

CRISPR genome engineering is imprecise – gene editing confirmation is critical

Genome edits can be thought of as a two-step process: a nuclease makes a cut at a specific sequence, and the resulting break is repaired either by non-homologous end joining (NHEJ) or through template-directed homology directed repair (HDR), both of which can be prone to imprecise gene editing.

Confirmation of the altered locus, whether engineering knock-ins, knock-outs or single nucleotide changes, to determine the extent of the modification or to verify that the specific change directed by the template was successful is therefore important.

Confidence through genomic confirmation

Phenotypic confirmation methods can be simple to perform and indicative of gene editing. However, they do not verify alterations at the genomic level. Therefore, downstream experimental findings are not definitive in nature as specific genetic changes cannot be attributed to an observed phenotype. Thus, many researchers seek to augment phenotypic confirmation with capillary electrophoresis-based Sanger sequencing.

Unexpected phenotypes explained through Sanger sequencing

Analysis of the sequence generated after an edit is especially important when the observed phenotype does not match the expectations of the edit. For example, a recent analysis of thyroid cancer used CRISPR editing to eliminate miR-146b, a hypothesized regulator of thyroid tumor cell growth (1).

Two different guide RNAs were used to try to eliminate the miR-146b locus. However, unexpectedly, two resulting cell lines showed reduced but not complete elimination of expression of miR-146b. The authors examined the engineered locus by Sanger sequencing and found deletions of 5 and 1 nucleotide in one of the cell lines, and a separate deletion of 2 nucleotides in the other cell line. The sequencing results showed that the locus was not completely eliminated, and the authors used this information to hypothesize other roles for miR-146b.

Sequencing the edited loci was able to provide an explanation for the unexpected phenotype.

CRISPR engineered animal models susceptible to misinterpretation

Zebrafish (Danio rerio) is used as an easy manipulatable model organism for vertebrate development. Before genome editing techniques were developed, researchers used small inhibiting RNAs (siRNAs) or morpholinos (chemical orthologs of nucleic acids) to knock down gene function. However, there were occasions when knock-out data obtained by genome edits failed to reproduce the expected loss-of-function phenotype. A review by Salanga and Salanga (2) points to several of these types of discrepancies and provides an explanation for them.

One example presented was that a CRISPR/Cas9 deletion of the pxr gene exon7-8, confirmed by Sanger sequencing, unexpectedly had no effect on the physiology being examined. Sequencing the mRNA transcript revealed a direct splicing from exon 6 to exon 9, completely bypassing the deletion (3).

Another example was that knockout of a gene (bag3), again confirmed by Sanger sequencing the edited locus, failed to produce cardiovascular defects and myofibrillar myopathy. They found that in the knockouts, a related gene (bag2) was significantly upregulated, compensating for the loss of bag3 (4).

And Zhu et al. found that deletions in the basement membrane glycoprotein Nidogen 1 (nid1a) were overcome by upregulating expression of structurally similar, but low sequence similarity, of other nidogen family genes (5). In each of these cases, reliance solely on phenotypic analysis would have led to misleading conclusions.

Adjacent confirmation sequencing unravels results

Sometimes genome edits produce detrimental “target adjacent” effects that confound the phenotypic analysis. Simkin et al. were investigating the induce pluripotent stem cells (iPSCs) from an epilepsy patient with a defined SNP mutation in the KCNQ2 gene (6). They hypothesized that by reverting the SNP to wild type using CRISPR genome engineering, they could restore firing frequency of neurons differentiated from the corrected cells.

They analyzed two different edited cell lines, and even though the identical edit was present, they had different phenotypic outcomes. Sequencing and other analysis of target adjacent regions revealed that the corrected clones were homozygous for SNPs and short intronic indels that flanked the edited locus. Surprisingly, these regions were heterozygous in the parent clone. This loss of heterozygosity (LOH) in this region, possibly through homologous recombination or allelic drop-out, complicated their interpretation of the electrophysiological phenotype.

Similar results were seen with other genes (6). These results show the importance of sequencing adjacent regions around an edit – changes in these sequences may affect the observed phenotype of the edited cells.

Genomic analysis identifies incomplete penetrance

Another challenge with correlating genotype to phenotype is that mutations can be incompletely penetrant and variably expressed. This means that even in a uniform genetic background, the phenotype can differ in the fraction of individuals affected, and the severity of the phenotype (7). This has been observed in human genetics, and in experiments where a large number of mouse knock-outs were constructed and analyzed (8).

Incomplete penetrance and variable expressivity can complicate the analysis of genome edits, since they can affect the way a phenotype presents to a researcher. Researchers should therefore ensure the designed edit is present and uniform in the cells or organism by sequencing the locus.

Sometimes genomic confirmation pinpoints translational impacts

Finally, another example of the importance of sequencing an edit comes from observations that certain mutations can affect the translation of unaffected genes, a phenomenon known as transcriptional adaptation (9). There are several different ways transcriptional adaptation can occur – including direct activation of heterologous gene expression, to altering mRNA decay pathways, to changes in epigenetic marks on DNA.

Transcriptional adaptation can override or alter the phenotypes produced by an edit. It is therefore important for an investigator to confirm the genotypic change at a locus before proceeding to interpret a phenotype.

Don’t allow faulty CRISPR editing to put your scientific insights at risk

The point of all these examples is that if the investigators had only examined the phenotype following a genome edit, they might have drawn incorrect conclusions – that the edit was unsuccessful, that the gene was not important for the phenotype, or that the phenotypes observed were strictly due to off-target events.

No matter what type of genomic alteration, cells will endeavor to “find a way” to overcome the detrimental effects of the mutational changes to survive. In order to correctly interpret the results, it is therefore critical for investigators to not rely on phenotypic characterization of a genome edit on its own – the underlying sequence should also be determined.

Confidently interpret gene editing results using simple data analysis

Sanger sequencing uses a simple workflow for highly accurate determination of the nucleic acid sequence with the single base resolution required for gene editing confirmation.

In addition, data analysis of knock-ins, knock-outs, deletions, insertions, or single nucleotide changes, is straightforward using the Applied Biosystems SeqScreener Gene Edit Confirmation (SGC) app. It’s a free software app that facilitates in-depth gene editing analysis by users of all skill levels, providing easy-to-interpret direct visualization of Sanger traces and graphical displays of editing efficiency, clone purity, straightforward, sample-to-answer workflow for highly accurate genomic-level gene editing confirmation.

Interested in performing Sanger sequencing? Learn more about:

References

De Santa Inez et al. Targeting the highly expressed microRNA miR-146b with CRISPR/Cas9 gene editing system in thyroid cancer. Int J Mol Sci 22:7992 (2021). https://doi.org/10.3390/ijms22157992
Slanaga CM and Salanga MC. Genotype to phenotype: CRISPR gene editing reveals genetic compensation as a mechanism for phenotypic disjunction of morphants and mutants. Int J Mol Sci 22:3472 (2021). https://doi.org/10.3390/ijms22073472
Salanga MC et al. CRISPR-Cas9-mutated pregnane X receptor (pxr) retains pregnenolone-induced expression of cyp3a65 in zebrafish (Danio rerio). Toxicol Sci 174:51-62 (2020). doi: 10.1093/toxsci/kfz246
Diofano F et al. Genetic compensation prevents myopathy and heart failure in an in vivo model of Bag3 deficiency. PLOS Genet 16:1-24 (2020). https://doi.org/10.1371/journal.pgen.1009088
Zhu P et al. Short body length phenotype is compensated by the upregulation of nidogen family members in a deleterious nid1a mutation of zebrafish. J Genet Genom 44:553-556 (2017). doi: 10.1016/j.jgg.2017.09.011
Simkin et al. Homozygous might be hemizygous: CRISPR/Cas9 editing in iPSCs results in detrimental on-target defects that escape standard quality controls. Stem Cell Reports 17:993-1008 (2022). https://doi.org/10.1016/j.stemcr.2022.02.008
Kingdom R and Wright CF. Incomplete penetrance and variable expressivity: From clinical studies to population cohorts. Frontiers in Genetics 13:920390 (2022). doi: 3389/fgene.2022.920390
Dickinson ME et al. High-throughput discovery of novel developmental phenotypes. Nature 537:508-514 (2016). https://doi.org/10.1038/nature19356.
Jakutis G and Stainier DYR. Geneotype-phenotype relationships in the context of transcriptional adaptation and genetic robustness. Ann Rev Genet 55:71-91 (2021). https://doi.org/10.1146/annurev-genet-071719-020342