The ability to selectively edit targeted genome regions using a technique known as CRISPR–Cas editing has transformed many areas of biological research. These advances have raised the question of whether this technique will be used in the clinic in future to treat or prevent disease. Clinical trials using this technique to edit human cells are already under way, for example testing the use of edited immune cells to treat cancer1. There is ongoing debate about the potential use of CRISPR–Cas to modify the human genome in an individual, a subject that raises many ethical and regulatory issues2. And if this were to occur, what extra scientific research would be necessary to reach the stage at which this approach could be implemented in the clinic? On page 413, Ma et al.3 report the use of CRISPR–Cas editing to repair a gene alteration associated with heart disease, in a study of human embryos grown in vitro. The authors thoroughly analyse the edited embryos, demonstrating that some technical hurdles often associated with using such genome-editing technology might be preventable.

The condition known as hypertrophic cardiomyopathy is an inherited cardiac disease. It can be caused by mutations in many different genes, including cardiac myosin-binding protein C (MYBPC3). This gene encodes a protein that contributes to the structural maintenance of heart muscle and the regulation of its contraction and relaxation4. The presence of one mutant copy of MYBPC3 causes symptoms that usually manifest as heart failure. Although existing treatments can lessen the symptoms, there is no way of tackling the underlying genetic cause in a patient5. Ma and colleagues investigated the use of gene editing to correct this mutation.

One option to prevent some mutation-associated conditions from being inherited is to use genetic testing during in vitro fertilization (IVF) fertility treatment. This enables the selection of embryos for implantation that do not contain the specific mutation. If one of the parents participating in IVF treatment had a mutant copy of MYBPC3, 50% of the couple's fertilized embryos would inherit the condition. Ma and colleagues propose a hypothetical scenario in which their approach could enable an individual with hypertrophic cardiomyopathy to increase the percentage of embryos available for IVF implantation that would not inherit the disease.

In the past few years, CRISPR–Cas has been developed6,7 to efficiently and precisely edit the human genome. Although this method has been widely adopted using mammalian cell-culture systems and animal embryo models, only three published studies8,9,10 report the use of this technique in human embryos.

The CRISPR–Cas editing system needs just two components to modify DNA: a guide RNA sequence and a Cas nuclease enzyme, with Cas9 the most commonly used. The specific genomic target is determined by the guide RNA, which forms a complex with Cas9, enabling the enzyme to target a genomic site that contains a matching sequence. This is where Cas9 cleaves the DNA, causing a double-stranded break6,7.

The formation of double-stranded breaks can activate one of the two main DNA-repair pathways in the cell: either the non-homologous end joining (NHEJ) pathway or the homology-directed repair (HDR) pathway. The NHEJ pathway repairs a break by randomly adding or deleting nucleotides, which results in changes to the DNA sequence, making it unsuitable for gene-correction purposes. Therefore, approaches focus on HDR, which uses homologous (matching) sequences to repair the DNA breaks, making it possible to introduce specific sequences to enable tailored repair.

Timed delivery of CRISPR components at an appropriate point in the cell's division cycle — either at the transition between the G1 and S phases or at the transition between G2 and M phases — might lead to preferential use of the HDR pathway11. Unfortunately, in previous reports, the efficiency of HDR repair after CRISPR–Cas9 action has been undesirably low both in cultured human embryonic stem cells (around 2% efficiency)11 and in the human embryo (14.3–25% efficiency)8.

Ma and colleagues created CRISPR–Cas9 gene-editing constructs to target MYBPC3, and verified and analysed the gene targeting using human stem cells. They then began to work with human embryos. One of the main challenges of using CRISPR–Cas9 to edit human-embryo genomes has been the phenomenon of mosaicism, in which gene-editing inefficiencies result in a developing embryo that has both edited and unedited cells8,9,10 (Fig. 1a). This might lead to a mixture of healthy and diseased cells in various tissues and organs, possibly causing disease symptoms.

Figure 1: Gene editing in embryos.
figure 1

a, The gene-editing system6,7 CRISPR–Cas can repair a mutation in a gene. In the three previous published studies8,9,10 using this technique in human embryos grown in vitro, an egg (oocyte) was fertilized by sperm and then the editing components were injected into the cell. These include the enzyme Cas9 and a guide RNA sequence that helps to direct the editing machinery to a specific location in the genome, for example to repair a mutation inherited from the father. However, editing is often inefficient, and later-stage embryos can contain a mixture of repaired and non-repaired cells — a phenomenon known as mosaicism. b, Ma et al.3 took an alternative approach to correct a mutation in the MYBPC3 gene (which is associated with heart disease) in human embryos grown in vitro. They injected gene-editing components and sperm into oocytes that contained non-mutated versions of MYBPC3. Half the sperm used had a MYBPC3 mutation. The oocytes were injected at the metaphase II stage of their cell cycle. Ma et al. report that 42 of 58 embryos tested (72.4%) did not have the MYBPC3 mutation, and their analysis suggests that the maternal copy of the gene is used as a template to guide the repair. This approach resulted in efficient, uniform gene editing in embryos, which progressed to reach a later stage of embryonic development.

Ma et al. investigated a situation in which the father had one mutant copy of MYBPC3 and the mother had only wild-type copies of the gene. In control experiments, 47.4% (9 out of 19) of embryos fertilized in vitro did not inherit a mutant copy of MYBPC3, the proportion expected given that half of the sperm should have the wild-type copy of the gene. The authors demonstrated that, if genome-editing components were injected together with sperm into a human egg (oocyte) at a stage in the oocyte cell cycle known as metaphase II (Fig. 1b), then 72.4% of the resulting embryos (42 out of 58) had only the wild-type version of MYBPC3. The other 16 embryos had copies of MYBPC3 with signs of NHEJ-mediated editing that targeted the mutant version of the gene but did not repair it to the wild-type version.

Previous studies of CRISPR–Cas editing in human embryos8,9,10 added the gene-editing components after fertilization. The low levels of mosaicism (one of the 42 embryos with wild-type MYBPC3 was a mosaic of cells that each had one of two versions of the wild-type gene) described by Ma et al. might be a result of the gene editing happening before the first cell division occurred.

The edited embryos developed similarly to the control embryos, with 50% reaching an early stage of development known as the blastocyst, in which the embryos contain different cell types. This indicates that editing does not block development.

When injecting the editing components, the authors also included a nucleotide sequence containing wild-type MYBPC3 that could be used as a repair template by the editing machinery. The authors designed the repair template to encode the same amino acids as the wild-type maternal copy of the gene, but used some different nucleotides so that they could distinguish between repair using either the maternal copy of MYBPC3 or the introduced repair template.

Interestingly, when the authors analysed the MYBPC3 gene correction in the embryos, they found that only one of the 42 embryos that had wild-type MYBPC3 had used the introduced template for repair. Their results indicate that the wild-type maternal copy of MYBPC3 probably provided the repair template, rather than the introduced copy. This contrasts with the authors' observations in stem cells, in which the introduced nucleotide template was used for the repair process. The authors propose that the DNA-repair mechanism operating in an early embryo differs from that occurring in stem cells. In the embryo in which correction using the repair template could be detected, the template sequence was found in only a subset of cells, and the remaining cells in this embryo had probably been corrected using the maternal version of the gene.

Another hurdle for the use of CRISPR–Cas9 technology is the possibility of off-target edits that might arise if editing components bind to genomic regions that have high similarity to the sequence targeted by the RNA guide. Ma et al. did not detect a type of genetic alteration associated with off-target edits in the wild-type copy of MYBPC3, suggesting high-fidelity editing.

Potential off-target changes in other genes were evaluated. Strikingly, the off-target testing did not detect any mutations in the sequencing results from the embryonic cells, leading the authors to conclude that the targeting was accurate. This is an important finding because off-target edits have been described as a challenge for the use of CRISPR technology in the clinic. However, because the targeted gene sequences would vary, and variations would also be present in the sequences of each genome being edited, the risk of off-target events might vary for each individual case.

The authors investigated a situation in which one copy of a gene needed to be targeted for editing, and found that genetic repair relied on the other wild-type copy of the gene. However, in some diseases, both maternal and paternal copies of a gene are mutant. In this situation, the wild-type copy is absent, so a repair strategy would have to rely on the use of an introduced template sequence. Inducing double-stranded breaks in both copies of the mutant genes might enable such a template to be used; however, it is conceivable that this might lead to increased usage of the deleterious NHEJ pathway. In such a situation, it is probable that specific strategies to prevent NHEJ, such as the use of blocking agents, would be crucial. That strategy would also be beneficial even when only one copy of a gene needs to be targeted, because 16 out of 58 of the embryos analysed by Ma and colleagues showed signs of NHEJ activity. However, there is a clear need to ensure that such strategies do not result in other damaging effects on a developing embryo and its genome.

Although Ma and colleagues' demonstration that embryonic genome integrity is maintained after CRISPR–Cas9 editing is promising, further studies and optimization of this technology will be needed. These will have to confirm that the approach is safe in terms of criteria such as mosaicism, off-target editing and the detection of abnormalities in edited embryos before it can be used as a therapy for inherited diseases. Nevertheless, this study is paving the way as part of investigations that might lead to CRISPR–Cas9 reaching the clinic in the future. Until then, embryo genetic testing during IVF remains the standard way to prevent the transmission of inherited diseases in human embryos.Footnote 1