Does gene editing have any technical cross-species limits?

Does gene editing have any technical cross-species limits?

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

How far could you go in cross-species gene editing?

Is it possible for example to introduce plant genes into human DNA and vice versa?

Question: How is gene editing limited by the "donor" and "target" species?

So, I'll assume that you are talking about gene editing in a laboratory (e.g. using a technique like CRISPR).

Theoretically, there are almost no limits to what you can do - or at least try. Fundamentally DNA is the same in all living (and non-living) things on earth, so you can absolutely take a gene from a plant and "paste" it into a human cell/genome (with some additional work needed to make the pasting work). You probably even go further and try to isolate whole chromosomes from one organism and (literally) inject them into cells of another one.

The problem with this radical approach to gene editing is, that it probably won't do much (good). A human cell likely doesn't recognise a plant gene as a gene and so just ignores it (it might still cause indirect problems for the cell though) and if you start making very drastic changes like adding chromosomes the chances that your cells will just die are super high.

This means that in practice you either need to (heavily) modify a gene you want to insert to match the structure and 'expectations' of the target organism - you basically have to make a humanised version of a given plant gene - or you can only use genes from similar enough organisms: as a general rule genes from the same kingdom (bacteria, archaea, plants, fungi, animals) are likely to retain some degree of function in another species (but how much exactly or even a general effect is almost impossible to predict).

What is CRISPR gene editing, and how does it work?

Merlin Crossley receives funding from the Australian Research Council and National Health and Medical Research Council. He is a Trustee of the Australian Museum, on the Board of the Australian Science Media Centre, and the Editorial Board of The Conversation.


UNSW provides funding as a member of The Conversation AU.

The Conversation UK receives funding from these organisations

You’ve probably read stories about new research using the gene editing technique CRISPR, also called CRISPR/Cas9. The scientific world is captivated by this revolutionary technology, since it is easier, cheaper and more efficient than previous strategies for modifying DNA.

The term CRISPR/Cas9 stands for Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR associated protein 9. The names reflect important features identified during its discovery, but don’t tell us much about how it works, as they were coined before anyone understood what it was.

The Pros of Gene Editing

Firstly, here are the advantages of genome editing technology.

1. Tackling and Defeating Diseases:

Most deadly and severe diseases in the world have resisted destruction. A number of genetic mutations that humans suffer will end only after we actively intervene and genetically engineer the next generation.

  • Cancer Therapeutics: New immunotherapy can be developed using genetic editing that can treat cancer. Modification of T-cells using CRISPR can locate and kill cancer cells.
  • Drug Research: Genetic makeup can potentially speed up the drug discovery process. Some of the drug makers are already incorporating CRISPR technology in drug research and discovery phase.
  • Inherent diseases: With genetic editing, the scientist can prevent the inherent disease from flowing to the offspring. Diabetes and cystic fibrosis can also be eliminated.

2. Extend Lifespan

Genome editing could extend the human lifespan. The human lifespan has already shot up by a number of years, and we are already living longer and longer. Genetic engineering could make our time on Earth even long. There are specific, common illnesses and diseases that can take hold later in life and can end up killing us earlier than necessary. Genetic editing can reverse the most fundamental reasons for the body’s natural decline on a cellular level. So it can drastically improve both the span and the quality of life later on.

3. Growth In Food Production and Its Quality:

Genetic engineering can design foods that can withstand harsh temperatures and are packed full of all the right nutrients. Additionally, it could also be the answer to meet the heavy food demands that are still not met in many countries. We may also increase the medicinal value of our food and introduce edible vaccines.

4. Pest Resilient Crops:

According to Jennifer Doudna, CRISPR pioneer, “genome editing can address pest and nutrition challenges facing agriculture”. Instead of using tons of insecticides and pesticides, we can protect our plan in a healthier way.

CRISPR alternatives

CRISPR-mediated genome editing has drawbacks, though. The PAM requirement limits target sequences. Cas9 is large, so its gene is difficult to deliver to cells via vectors such as adeno-associated viruses commonly used in gene therapy. Scientists worry about off-target effects, although experts note that concerns about unintended mutations are often based on calculations from studies on improving editing. These studies may deliberately use low-specificity conditions to facilitate monitoring progress.

To ensure the highest confidence in their products, companies invest time and money in custom genome-editing methods focused on efficiency and specificity. Initial investments pay off, industry scientists say, by preventing problems later in development.

ZFNs are the genome-editing reagents used by the genomic medicine company Sangamo, based in Brisbane, California. Chief Technology Officer Ed Rebar explains that Sangamo’s core editing reagent is a ZFN dimer. The typical target site is 36 basepairs. Each ZFN is a chimeric protein of the nuclease domain from the FokI restriction enzyme and an array of zinc-finger DNA-binding domains built by “mixing and matching” from Sangamo’s archive of thousands of two-finger subunits. Strategies for diversifying the ZFN architecture for high targeting capability include attaching the FokI domain to the N- or C-terminus of the zinc-finger array and inserting base-skipping linkers between fingers. With Sangamo’s high-throughput, automated process for generating ZFNs, Rebar says, “Starting from a target gene name, we can generate an initial set of editing reagents within two weeks.”

In a demonstration study, Rebar and colleagues designed ZFNs that introduced indels at 25 of 28 bases in a promoter relevant to studying hemoglobinopathies (3). Despite this precision and the advantage of being smaller than Cas9, ZFNs are not as commonly used as CRISPR-based methods. Sangamo provides ZFNs via industry and academic partnerships but holds the modules, expertise—and patents—for making them.

TALENs attach FokI to arrays of DNA-binding modules, originally from plant pathogens, that each target a single basepair. TALENs are smaller than Cas9, but larger than ZFNs. The modules have high DNA-binding affinity but include repeated sequences that create cloning challenges.

Dan Carlson is chief scientific officer at Recombinetics, a St. Paul, Minnesota–based biotechnology company that uses TALENs and CRISPR to generate animals and cell lines for clinical research models and agriculture. Using these methods, he says, “we can target almost any site in a genome.” With in-house resources, even TALENs take only “a few hundred bucks and about a week” to generate, Carlson adds, so scientists choose the method that is most reproducible, consistent, and specific, based on pilot studies. These initial investments ensure the company is responsible with resources, he says. “It costs too much to sort out problems on the back end.”

Meganucleases, also called homing endonucleases, are smaller than Cas9, despite their name, which refers to recognition sequences that can be up to 40 basepairs in length. Hybrid megaTALs combine the simple assembly of TALENS with the DNA-cleavage specificity of meganucleases. Two biotech companies that use meganuclease-based methods are Bluebird Bio in Cambridge, Massachusetts, and Precision BioSciences in Durham, North Carolina.

Barry Stoddard, a structural biologist at Fred Hutchinson Cancer Research Center, Seattle, has a panel of 50–60 meganucleases that his lab engineers to recognize specific sequences. “It takes one day to make CRISPR to target a gene,” he says, “and 100 days to make a meganuclease.” Still, Stoddard gets many requests for engineered meganucleases, because their precision is highly valued for applications such as developing therapeutics for which “100 days is nothing.”

CRISPR/Cas9 Mediated Gene Editing

Pioneering Discoveries in CRISPR/Cas9 Technology

The bacterial CRISPR locus was first described by Francisco Mojica (23) and later identified as a key element in the adaptive immune system in prokaryotes (24). The locus consists of snippets of viral or plasmid DNA that previously infected the microbe (later termed “spacers”), which were found between an array of short palindromic repeat sequences. Later, Alexander Bolotin discovered the Cas9 protein in Streptococcus thermophilus, which unlike other known Cas genes, Cas9 was a large gene that encoded for a single-effector protein with nuclease activity (25). They further noted a common sequence in the target DNA adjacent to the spacer, later known as the protospacer adjacent motif (PAM)—the sequence needed for Cas9 to recognize and bind its target DNA (25). Later studies reported that spacers were transcribed to CRISPR RNAs (crRNAs) that guide the Cas proteins to the target site of DNA (26). Following studies discovered the trans-activating CRISPR RNA (tracrRNA), which forms a duplex with crRNA that together guide Cas9 to its target DNA (27). The potential use of this system was simplified by introducing a synthetic combined crRNA and tracrRNA construct called a single-guide RNA (sgRNA) (28). This was followed by studies demonstrating successful genome editing by CRISPR/Cas9 in mammalian cells, thereby opening the possibility of implementing CRISPR/Cas9 in gene therapy (29) ( Figure 1 ).

Hallmarks of CRISPR Gene Therapy. Timeline highlighting major events of traditional gene therapy, CRISPR development, and CRISPR gene therapy. The text in red denotes gene therapy events which have raised significant ethical concerns.

Mechanistic Overview of CRISPR/Cas9-Mediated Genome Editing

CRISPR/Cas9 is a simple two-component system used for effective targeted gene editing. The first component is the single-effector Cas9 protein, which contains the endonuclease domains RuvC and HNH. RuvC cleaves the DNA strand non-complementary to the spacer sequence and HNH cleaves the complementary strand. Together, these domains generate double-stranded breaks (DSBs) in the target DNA. The second component of effective targeted gene editing is a single guide RNA (sgRNA) carrying a scaffold sequence which enables its anchoring to Cas9 and a 20 base pair spacer sequence complementary to the target gene and adjacent to the PAM sequence. This sgRNA guides the CRISPR/Cas9 complex to its intended genomic location. The editing system then relies on either of two endogenous DNA repair pathways: non-homologous end-joining (NHEJ) or homology-directed repair (HDR) ( Figure 2 ). NHEJ occurs much more frequently in most cell types and involves random insertion and deletion of base pairs, or indels, at the cut site. This error-prone mechanism usually results in frameshift mutations, often creating a premature stop codon and/or a non-functional polypeptide. This pathway has been particularly useful in genetic knock-out experiments and functional genomic CRISPR screens, but it can also be useful in the clinic in the context where gene disruption provides a therapeutic opportunity. The other pathway, which is especially appealing to exploit for clinical purposes, is the error-free HDR pathway. This pathway involves using the homologous region of the unedited DNA strand as a template to correct the damaged DNA, resulting in error-free repair. Experimentally, this pathway can be exploited by providing an exogenous donor template with the CRISPR/Cas9 machinery to facilitate the desired edit into the genome (30).

CRISPR/Cas9 mediated gene editing. Cas9 in complex with the sgRNA targets the respective gene and creates DSBs near the PAM region. DNA damage repair proceeds either through the NHEJ pathway or HDR. In the NHEJ pathway, random insertions and deletions (indels) are introduced at the cut side and ligated resulting in error-prone repair. In the HDR pathway, the homologous chromosomal DNA serves as a template for the damaged DNA during repair, resulting in error-free repair.

Materials and methods

Data and code availability

The source code for SAMap is publicly available at Github ( copy archived at swh:1:rev:c696585f8fe41ec1599b0720df579f3cb14f935b Tarashansky et al., 2021), along with the code to perform the analysis and generate the types of plots presented in the figures. We also provide a wrapper function to launch a graphical user interface provided by the SAM package to interactively explore both datasets in the combined manifold. The datasets analyzed in this study are detailed in Supplementary file 1 with their accessions and annotations provided.

The SAMap algorithm

The SAMap algorithm contains three major steps: preprocessing, mutual nearest neighborhood alignment, and gene-gene correlation initialization. The latter two are repeated for three iterations, by default, to balance alignment performance and computational runtime.


Generate gene homology graph via reciprocal BLAST

We first construct a gene-gene bipartite graph between two species by performing reciprocal BLAST of their respective transcriptomes using tblastx, or proteomes using blastp. tblastn and blastx are used for BLAST between proteome and transcriptome. When a pair of genes share multiple High Scoring Pairs (HSPs), which are local regions of matching sequences, we use the HSP with the highest bit score to measure homology. Only pairs with E-value <10 −6 are included in the graph.

Although we define similarity using BLAST, SAMap is compatible with other protein homology detection methods (e.g. HMMER [Eddy, 2008]) or orthology inference tools (e.g. OrthoClust [Yan et al., 2014] and eggNOG [Huerta-Cepas et al., 2019]). While each of these methods has known strengths and limitations, BLAST is chosen for its broad usage, technical convenience, and compatibility with low-quality transcriptomes.

We encode the BLAST results into two triangular adjacency matrices, A and B , each containing bit scores in one BLAST direction. We combine A and B to form a gene-gene adjacency matrix G . After symmetrizing G , we remove edges that only appear in one direction: G = R e c i p ( 1 2 A + B + A + B T ) ∈ R m 1 + m 2 × m 1 + m 2 , where R e c i p only keeps reciprocal edges, and m 1 and m 2 are the number of genes of the two species, respectively. To filter out relatively weak homologies, we also remove edges where G a b < 0.25 m a x b ( G a b ) . Edge weights are then normalized by the maximum edge weight for each gene and transformed by a hyperbolic tangent function to increase discriminatory power between low and high edge weights, G ^ a b = 0.5 + 0.5 t a n h ( 10 G a b / m a x b ( G a b ) - 5 ) .

Construct manifolds for each cell atlas separately using the SAM algorithm

The single-cell RNAseq datasets are normalized such that each cell has a total number of raw counts equal to the median size of single-cell libraries. Gene expressions are then log-normalized with the addition of a pseudocount of 1. Genes expressed (i.e. l o g 2 ( D + 1 ) > 1 ) in greater than 96% of cells are filtered out. SAM is run using the following parameters: preprocessing = ‘StandardScaler’, weight_PCs = False, k = 20, and npcs = 150. A detailed description of parameters is provided previously (Tarashansky et al., 2019). SAM outputs N 1 and N 2 , which are directed adjacency matrices that encode k-nearest neighbor graphs for the two datasets, respectively.

SAM only includes the top 3000 genes ranked by SAM weights and the first 150 principal components (PCs) in the default mode to reduce computational complexity. However, downstream mapping requires PC loadings for all genes. Thus, in the final iteration of SAM, we run PCA on all genes and take the top 300 PCs. This step generates a loading matrix for each species i , L i ∈ R 300 × m i .

Mutual nearest neighborhood alignment

Transform feature spaces between species

For the gene expression matrices Z i ∈ R n i × m i , where n and m are the number of cells and genes respectively, we first zero the expression of genes that do not have an edge in G ^ and standardize the expression matrices such that each gene has zero mean and unit variance, yielding Z

i . G ^ represents a bipartite graph in the form of G ^ = 0 m 1 , m 1 H ∈ R m 1 × m 2 H T ∈ R m 2 × m 1 0 m 2 , m 2 , where 0 m , m is m × m zero matrix and H is the biadjacency matrix. Letting H 1 = H and H 2 = H T encoding directed edges from species 1 to 2 and 2 to 1, respectively, we normalize the biadjacency matrix H i such that each row sums to 1: H ^ i = S u m N o r m ( H i ) ∈ R m i × m j , where the S u m N o r m function normalizes the rows to sum to 1. The feature spaces can be transformed between the two species via weighted averaging of gene expression, Z

Project single-cell gene expressions into a joint PC space

We project the expression data from two species into a joint PC space (Barkas et al., 2019), P i = Z

i j L j T . We then horizontally concatenate the principal components P i and P i j to form P ^ i ∈ R n i × 600 .

Calculate k-nearest cross-species neighbors for all cells

Using the joint PCs, P ^ i , we identify for each cell the k -nearest neighbors in the other dataset using cosine similarity ( k = 20 by default). Neighbors are identified using the hnswlib library, a fast approximate nearest-neighbor search algorithm (Malkov and Yashunin, 2020). This outputs two directed biadjacency matrices C i ∈ R n i × n j for ( i , j ) = ( 1 , 2 ) or ( 2 , 1 ) with edge weights equal to the cosine similarity between the PCs.

Apply the graph-coarsening mapping kernel to identify cross-species mutual nearest neighborhoods

To increase the stringency and confidence of mapping, we only rely on cells that are mutual nearest cross-species neighbors, which are typically defined as two cells reciprocally connected to one another (Haghverdi et al., 2018). However, due to the noise in cell-cell correlations and stochasticity in the kNN algorithms, cross-species neighbors are often randomly assigned from a pool of cells that appear equally similar, decreasing the likelihood of mutual connectivity between individual cells even if they have similar expression profiles. To overcome this limitation, we integrate information from each cell’s local neighborhood to establish more robust mutual connectivity between cells across species. Two cells are thus defined as mutual nearest cross-species neighbors when their respective neighborhoods have mutual connectivity.

Specifically, the nearest neighbor graphs N i generated by SAM are used to calculate the neighbors of cells t i hops away along outgoing edges: N ̄ i = N i t i , where N ̄ i are adjacency matrices that contain the number of paths connecting two cells t i hops away, for i = 1 or 2. t i determines the length-scale over which we integrate incoming edges for species i . Its default value is 2 if the dataset size is less than 20,000 cells and 3 otherwise. However, cells within tight clusters may have spurious edges connecting to other parts of the manifold only a few hops away. To avoid integrating neighborhood information outside this local structure, we use the Leiden algorithm (Traag et al., 2019) to cluster the graph and identify a local neighborhood size for each cell (the resolution parameter is set to 3 by default). If cell a belongs to cluster c a , then its neighborhood size is l a = | c a | . For each row a in N ̄ i we only keep the l a geodesically closest cells, letting the pruned graph update N ^ i .

Edges outgoing from cell a i in species i are encoded in the corresponding row in the adjacency matrix: C i , a i . We compute the fraction of the outgoing edges from each cell that target the local neighborhood of a cell in the other species: C

i , a i b j = ∑ c ∈ X j , b j C i , a i c , where X j , b j is the set of cells in the neighborhood of cell b j in species j and C

i , a i b j is the fraction of outgoing edges from cell a i in species i targeting the neighborhood of cell b j in species j .

To reduce the density of C

i so as to satisfy computational memory constraints, we remove edges with weight less than 0.1. Finally, we apply the mutual nearest neighborhood criterion by taking the element-wise, geometric mean of the two directed bipartite graphs: C

2 . This operation ensures that only bidirectional edges are preserved, as small edge weights in either direction results in small geometric means.

Assign the k-nearest cross-species neighborhoods for each cell

Given the mutual nearest neighborhoods C

∈ R n 1 × n 2 , we select the k nearest neighborhoods for each cell in both directions to update the directed biadjacency matrices C 1 and C 2 : C 1 = K N N ( C

T , k ) , with k = 20 by default.

Stitch the manifolds

We use C 1 and C 2 to combine the manifolds N 1 and N 2 into a unified graph. We first weight the edges in N 1 and N 2 to account for the number of shared cross-species neighbors by computing the one-mode projections of C 1 and C 2 . In addition, for cells with strong cross-species alignment, we attenuate the weight of their within-species edges. For cells with little to no cross-species alignment, their within-species are kept the same to ensure that the local topological information around cells with no alignment is preserved.

Specifically, we use N 1 and N 2 to mask the edges in the one-mode projections, N

1 = U ( N 1 ) ∘ ( N o r m ( C 1 ) N o r m ( C 2 ) ) and N

2 = U ( N 2 ) ∘ ( N o r m ( C 2 ) N o r m ( C 1 ) ) , where U ( E ) sets all edge weights in graph E to 1 and N o r m normalizes the outgoing edges from each cell to sum to 1. The minimum edge weight is set to be 0.3 to ensure that neighbors in the original manifolds with no shared cross-species neighbors still retain connectivity: N ∼ 1 , i j = m i n ( 0.3 , N ∼ 1 , i j ) and N ∼ 2 , i j = m i n ( 0.3 , N ∼ 2 , i j ) for all edges ( i , j ) . We then scale the within-species edges from cell i by the total weight of its cross-species edges: N

1 , i = ( 1 - 1 k ∑ j = 1 n 2 C 1 , i j ) N

2 , i = ( 1 - 1 k ∑ j = 1 n 1 C 2 , i j ) N

2 , i . Finally, the within- and cross-species graphs are stitched together to form the combined nearest neighbor graph N : N = [ N

2 ] . The overall alignment score between species 1 and 2 is defined as S = 1 n 1 + n 2 ( ∑ i = 1 n 1 ∑ j = 1 n 2 C 1 , i j + ∑ i = 1 n 2 ∑ j = 1 n 1 C 2 , i j ) .

Homology graph refinement

Update edge weights in the gene-gene bipartite graph with expression correlations

To compute correlations between gene pairs, we first transfer expressions from one species to the other: Z ̄ i , n i m j = C i , n i Z j , m j , where Z ̄ i , n i m j is the imputed expressions of gene m j from species j for cell n i in species i , and C i , n i is row n i of the biadjacency matrix encoding the cross-species neighbors of cell n i in species i , all for ( i , j ) = ( 1,2 ) and ( 2 , 1 ) . We similarly use the manifolds constructed by SAM to smooth the within-species gene expressions using kNN averaging: Z ̄ j , m j = N j , m j Z j , m j , where N j is the nearest-neighbor graph for species j . We then concatenate the within- and cross-species gene expressions such that the expression of gene m j from species j in both species is Z ̄ m j = Z ̄ i , m j ⊕ Z ̄ j , m j .

For all gene pairs in the initial unpruned homology graph, G ^ , we compute their correlations, G ^ a b : = θ ( 0 ) C o r r ( Z ̄ a , Z ̄ b ) , where θ ( 0 ) is a Heaviside step function centered at 0 to set negative correlations to zero. We then use the expression correlations to update the corresponding edge weights in G ^ , which are again normalized through G ^ a b = 0.5 + 0.5 t a n h ( 10 G ^ a b / m a x b ( G ^ a b ) - 5 ) .

Annotation of cell atlases

To annotate the primary zebrafish and Xenopus cell types, the cell subtype annotations provided by the original publications (Briggs et al., 2018 Wagner et al., 2018) are coarsened using a combination of the manual matching and developmental hierarchies. For example, as ‘involuting marginal zone’ in Xenopus is manually matched to ‘non-dorsal margin’, 'dorsal margin' 'non-dorsal margin involuted', and ‘dorsal margin involuted’ in zebrafish, we label these cells as ‘involuting marginal zone’. In cases where the matching is insufficient to coarsen the annotations, we use the provided developmental trees to name a group of terminal cell subtypes by their common ontogenic ancestor. Cell types that do not cluster well in the manifold reconstructed by SAM are excluded from the comparison. These include germline, heart, and olfactory placode cells, as they are mixed with other cell types in the Xenopus atlas. The germline cells are scattered across the reconstructed manifold and do not concentrate in a distinct cluster. The heart cells and olfactory placode cells are inextricably mixed with larger populations of intermediate mesoderm and placodal cells, respectively. Similarly, the iridoblast, epiphysis, nanog + , apoptotic-like, and forerunner cells are excluded because they do not cluster distinctly in the zebrafish atlas.

The annotations provided by their respective studies are used to label the cells in the Spongilla, Hydra, planarian, and mouse atlases. To annotate the schistosome cells, we use known marker genes to annotate the main schistosome tissue types (Li et al., 2021). Annotations for all single cells in all datasets are provided in Supplementary file 1.


The combined manifold N is embedded into 2D projections using UMAP implemented in the scanpy package (Wolf et al., 2018) by with the parameter min_dist = 0.1. The sankeyD3 package ( in R is used to generate the sankey plots. Edge thickness corresponds to the alignment score between mapped cell types. The alignment score between cell types a and b is defined as s a b = 1 | c a | + | c b | ( ∑ i ∈ c a ∑ j ∈ c b C 1 , i j + ∑ i ∈ c b ∑ j ∈ c a C 2 , i j ) , where c a and c b are the set of cells in cell types a and b , respectively. Cell type pairs with alignment score less than z are filtered out. By default, z is set to be 0.1.

The network graphs in Figure 5C are generated using the networkx package ( in python. To focus on densely connected cell type groups, we filter out cell type pairs with alignment score less than 0.05.

Identification of gene pairs that drive cell type mappings

We define g 1 and g 2 to contain SAMap-linked genes from species 1 and 2, respectively. Note that a gene may appear multiple times as SAMap allows for one-to-many homology. Let X a 1 b 2 denote the set of all cells with cross species edges between cell types a 1 and b 2 . We calculate the average standardized expression of all cells from species i that are in X a 1 b 2 : Y i , g i = 1 | < x , x ∈ X a 1 b 2 >| ∑ x ∈ X a 1 b 2 Z

i , x , g i ∈ R | g i | is the standardized expression of genes g i in cell x . The correlation between Y 1 , g 1 and Y 2 , g 2 can be written as C o r r ( Y 1 , g 1 , Y 2 , g 2 ) = ∑ j = 1 | g 1 | S ( Y 1 , g 1 ) j ∘ S ( Y 2 , g 2 ) j , where S ( Z ) standardizes vector Z to have zero mean and unit variance. We use the summand to identify gene pairs that contribute most positively to the correlation. We assign each gene pair a score: h g = T ( S ( Y 1 , g 1 ) ) ∘ T ( S ( Y 2 , g 2 ) ) , where T ( Z ) sets negative values in vector Z to zero in order to ignore lowly-expressed genes. To be inclusive, we begin with the top 1000 gene pairs according to h g and filter out gene pairs in which one or both of the genes are not differentially expressed in their respective cell types (p-value > 10 −2 ), have less than 0.2 SAM weight, or are expressed in fewer than 5% of the cells in the cluster. The differential expression of each gene in each cell type is calculated using the Wilcoxon rank-sum test implemented in the scanpy function

Orthology group assignment

We use the eggNOG mapper (v5.0) (Huerta-Cepas et al., 2019) to assign each gene to an orthology group with default parameters. For the zebrafish-to-Xenopus mapping, genes are considered orthologs if they map to the same vertebrate orthology group. For the pan-species analysis, we group genes from all species with overlapping orthology assignments. In Figure 6B, each column corresponds to one of these groups. As each group may contain multiple genes from each species, we present the expression of the gene with the highest enrichment score per species. All gene names and corresponding orthology groups are reported in Supplementary file 5.

Paralog substitution analysis

SAMap outputs gene-gene correlations across the combined manifold for all pairs of genes in the homology graph. As determined by eggNOG, genes that map to the same orthology group for the two species’ most recent common ancestor are considered orthologs, and those that map to the same orthology group more ancestral than Vertebrata are considered as paralogs. We note that as eggNOG does not provide an orthology group corresponding to the osteichtyan ancestor, our analysis does not include the paralogs that duplicated in between the osteichtyan and the vertebrate ancestors. If a gene has significantly higher correlation to one of its paralogs than its ortholog (>0.3 by default), we consider its ortholog to have been substituted. Paralog substitutions are identified using the samap.analysis.ParalogSubstitutions function provided by the SAMap package.

The evolutionary time period in which paralogs were duplicated can be inferred by identifying their most recent shared orthology group. We calculate the enrichment of paralog substitutions for each taxonomic level (i.e. Chordata, Bilateria, Metazoa, Opisthokonta, and Eukaryota) using the eggNOG orthology group assignments. We normalize the number of substituting paralogs by the total number of paralogs at each level to calculate the rate of paralog substitution across evolutionary time.

To quantify the enrichment of substituting paralogs in each cell type, we define a cell type-specific substitution score. We first assign paralog substitution events to cell types if the paralogous gene pairs are enriched in any of their mappings. Each cell type k then has a set of substituting paralogs P k . The score S k for cell type k is calculated as S k = ∑ i ∈ P k 1 - n i m k , where n i is the number of paralogs of ortholog i normalized by the maximum number of paralogs observed across all genes to accounts for the fact that genes with more paralogs are more likely to match with substituting paralogs by random chance, and m k is the number of differentially expressed genes in cell type k . Similarly, the denominator accounts for the fact that cell types with more differentially expressed genes are more likely to have paralog substitutions by random chance. The substitution scores for cell types with annotated homologs across species are averaged.

Phylogenetic reconstruction of gene trees

We generate gene trees to validate the identity of genes involved in putative examples of paralog substitution and of Fox and Csrp transcriptional regulators that are identified as enriched in contractile cells. For this, we first gather protein sequences from potential homologs using the eggnog version 5.0 orthology database (Huerta-Cepas et al., 2019). For the Fox and Csrp phylogenies, we include all Fox clade I (Larroux et al., 2008) and Csrp/Crip homologs, respectively, from the seven species included in our study.

Alignment of protein sequences is performed with Clustal Omega version 1.2.4 using default settings as implemented on the EMBL EBI web services platform (Madeira et al., 2019). Maximum likelihood tree reconstruction is performed using IQ-TREE version 1.6.12 (Nguyen et al., 2015) with the ModelFinder Plus option (Kalyaanamoorthy et al., 2017). For the Csrp tree, we perform 1000 nonparametric bootstrap replicates to assess node support. For Fox, we utilize the ultrafast bootstrap support option with 1000 replicates. For each gene tree we choose the model that minimizes the Bayesian Information Criterion (BIC) score in ModelFinder. This results in selection of the following models: DCMut+R4 (Csrp) and VT+F + R5 (Fox). The final consensus trees are visualized and rendered using the ETE3 v3.1.1 python toolkit (Huerta-Cepas et al., 2016) and the Interactive Tree of Life v4 (Letunic and Bork, 2019).

KOG functional annotation and enrichment analysis

Using the eggNOG mapper, KOG functional annotations are transferred to individual transcripts from their assigned orthology group. For enrichment analysis, all genes enriched in the set of cell type pairs of interest are lumped to form the target set for each species. For example, the target set for Spongilla archaeocytes used in Figure 6C is composed of all genes enriched between Spongilla archaeocytes and other invertebrate stem cells. Note that this set includes genes from other species that are linked by SAMap to the Spongilla archeocyte genes. We include genes from other species in the target set to account for differences in KOG functional annotation coverage between species. As such, the annotated transcripts from all seven species are combined to form the background set. We use a hypergeometric statistical test (Eden et al., 2009) to measure the enrichment of the KOG terms in the target genes compared to the background genes.

Mapping zebrafish and Xenopus atlases using existing methods

For benchmarking, we use vertebrate orthologs as determined by eggNOG as input to Harmony (Korsunsky et al., 2019), LIGER (Welch et al., 2019), Seurat (Stuart et al., 2019), Scanorama (Hie et al., 2019), BBKNN (Polański et al., 2019), which are all run with default parameters. One-to-one orthologs are selected from one-to-many and many-to-many orthologs by using the bipartite maximum weight matching algorithm implemented in networkx. When using the one-to-one orthologs as input for SAMap, we run for only one iteration. The resulting integrated lower-dimensional coordinates (PCs for Seurat, Harmony, and Scanorama and non-negative matrix factorization coordinates for LIGER) and stitched graphs (BBKNN and SAMap) are all projected into 2D with UMAP (Figure 2—figure supplement 1A). The integrated coordinates are used to generate a nearest neighbor graph using the correlation distance metric, which is then used to compute the alignment scores in Figure 2—figure supplement 1B. The alignment scores for SAMap and BBKNN are directly computed from their combined graphs.

In situ hybridization in schistosomes

S. mansoni (strain: NMRI) juveniles are retrieved from infected female Swiss Webster mice (NR-21963) at

3 weeks post-infection by hepatic portal vein perfusion using 37°C DMEM supplemented with 5% heat inactivated FBS. The infected mice are provided by the NIAID Schistosomiasis Resource Center for distribution through BEI Resources, NIH-NIAID Contract HHSN272201000005I. In adherence to the Animal Welfare Act and the Public Health Service Policy on Humane Care and Use of Laboratory Animals, all experiments with and care of mice are performed in accordance with protocols approved by the Institutional Animal Care and Use Committees (IACUC) of Stanford University (protocol approval number 30366). In situ hybridization experiments are performed as described previously (Tarashansky et al., 2019), using riboprobes synthesized from gene fragments cloned with the listed primers: collagen (Smp_170340): GGTGAAGAAGGCTGTTGTGG , ACGATCCCCTTTCACTCCTG tropomyosin (Smp_031770): AAGCTGAAGTCGCCTCACTA , CATATGCCTCTTCACGCTGG troponin (Smp_018250): CGTAAACCTGGTCAGAAGCG , ATCCTTTTCCTCCAGAGCGT myosin regulatory light chain (Smp_132670): GAGACAGCGAGTAGTGGAGG , TGCCTTCTTTGATTGGAGCT wnt11 (Smp_156540): TGTGGTGATGAAGATGGCAG , CCACGGCCACAACACATATT frizzled (Smp_174350): CGAACAGGCGCATGACAATA , TGCTAGTCCTGTTGTCGTGT .


Genome editing technology provides revolutionary ways to change, regulate, determine and imagine genomes in large animals, potentially offering novel applications in biomedicine and agriculture. We anticipate greater numbers of applications materializing in the near future, such as genome-edited NHPs combined with SCNT, pig organ xenotransplantation used in clinical trials, and genome-edited livestock-derived meat making its way to the food table. However, challenges still remain with integrating genome editing into biomedicine and agriculture. It is obvious that the effectiveness and specificity of genome editing with currently available tools still needs improvement, and that the safety and ethical concerns of using genetically modified tissues, organs and animals remain a focus of considerable debate. Specifically, a strategy to generate large founder animals with a desired allele in one step, without a prolonged period of breeding, is in high demand. However, mosaic mutations, which are commonly observed in zygote injection-based genome editing, are another potential challenge in the editing of large animals. These issues could potentially be solved by tagging Cas9 with ubiquitin-proteasomal degradation signals [ 150] and introducing editing components in an appropriate format (i.e. a Cas9 protein/sgRNA complex) into very early-stage zygotes [ 151, 152]. Other possible strategies to reduce mosaicism have been discussed in a recent review [ 153].

Gene therapy, using genetic modification with exogenous DNA to treat inherited human diseases, offers new treatment modalities in multiple medical fields. Currently, three gene therapy products have been approved by the US Food and Drug Administration (FDA): LUXTURNA™ (manufactured by Spark Therapeutics, Inc.) for the treatment of patients with confirmed biallelic RPE65 mutation-associated retinal dystrophy, KYMRIAH™ (manufactured by Novartis Pharmaceuticals Corporation) for the treatment of patients up to 25 years of age with B-cell precursor acute lymphoblastic leukemia (ALL) that is refractory or in second or later relapse, and YESCARTA™ (manufactured by Kite Pharma, Inc.) for the treatment of adult patients with relapsed or refractory large B-cell lymphoma after two or more lines of systemic therapy. In addition, the European Medicines Agency has approved Glybera for lipoprotein lipase deficiency and the FDA has assigned LentiGlobin BB305 as a breakthrough therapy designation request for treatment of transfusion-dependent patients with β-thalassemia major. However, a broader spectrum of somatic cell editing needs to be developed, both ex vivo and in humans. Animal models that match and recapitulate the characteristics of human disease are a top priority for evaluating the efficacy and safety of gene therapy or cell-based therapy to treat human disease.

In spite of the substantial potential of genome editing for clinical and agricultural applications, safety and ethical issues cannot be ignored. Xenotransplantation provides hope to patients living with organ failure and waiting for a donor, yet the use of animal organs and tissues in humans is still not fully accepted due to safety and ethical concerns. Further confirmation of the efficacy and safety of xenotransplantation is urgently needed for the procedure to gain acceptance. On the other hand, with the advent of interspecies chimeras using blastocyst complementation, human organs may one day be produced in large animals. However, researchers and the public still have concerns about the risk of human cells integrating into the host animal's brain or germline, and these concerns need to be taken into account. Unlike transgene technology, which introduces an exogenous gene into the host genome randomly, genome editing only changes the endogenous gene in an efficient and accurate way. With these new technologies, the FDA is maintaining a product-focused, science-based regulatory policy with specific legal standards applied to different types of products. The FDA has determined that animals with intentionally altered genomes should be subjected to regulations under the provisions of new animal drugs [ 154]. Unlike the FDA, the US Department of Agriculture (USDA) has stated that the USDA will not regulate genetically modified plants produced by the new genome editing techniques [ 155], which will definitely accelerate the commercialization of genome-edited organisms. With further studies to solve the ‘off-target’ effects and potential risks to the host genome, genome editing of animals may become more accepted by the public.

In summary, the rapid progress of genome editing in large animals has resulted in the production of many valuable animals for human disease models, xenotransplantation and the agricultural economy (Table 1). Further optimization of the existing genome editing system and the generation of new tools for precise gene modification will additionally accelerate the development of genetically modified animals, organs and tissues for agriculture, regenerative medicine and therapeutic applications.

Is gene editing ethical?

If you bring up the subject of gene editing, the debate is sure to become heated. But are we slowly warming to the idea of using gene editing to cure genetic diseases, or even create “designer babies?”

Share on Pinterest Will gene editing become a part of everyday medicine?

Gene editing holds the key to preventing or treating debilitating genetic diseases, giving hope to millions of people around the world. Yet the same technology could unlock the path to designing our future children, enhancing their genome by selecting desirable traits such as height, eye color, and intelligence.

While gene editing has been used in laboratory experiments on individual cells and in animal studies for decades, 2015 saw the first report of modified human embryos.

The number of published studies now stands at eight, with the latest research having investigated how a certain gene affects development in the early embryo and how to fix a genetic defect that causes a blood disorder .

The fact that gene editing is possible in human embryos has opened a Pandora’s box of ethical issues.

So, who is in favor of gene editing? Do geneticists feel differently about this issue? And are we likely to see the technology in mainstream medicine any time soon?

Gene editing is the modification of DNA sequences in living cells. What that means in reality is that researchers can either add mutations or substitute genes in cells or organisms.

While this concept is not new, a real breakthrough came 5 years ago when several scientists saw the potential of a system called CRISPR/Cas9 to edit the human genome.

CRISPR/Cas9 allows us to target specific locations in the genome with much more precision than previous techniques. This process allows a faulty gene to be replaced with a non-faulty copy, making this technology attractive to those looking to cure genetic diseases.

The technology is not foolproof, however. Scientists have been modifying genes for decades, but there are always trade-offs. We have yet to develop a technique that works 100 percent and doesn’t lead to unwanted and uncontrollable mutations in other locations in the genome.

In a laboratory experiment, these so-called off-target effects are not the end of the world. But when it comes to gene editing in humans, this is a major stumbling block.

Here, the ethical debate around gene editing really gets off the ground.

When gene editing is used in embryos — or earlier, on the sperm or egg of carriers of genetic mutations — it is called germline gene editing. The big issue here is that it affects both the individual receiving the treatment and their future children.

This is a potential game-changer as it implies that we may be able to change the genetic makeup of entire generations on a permanent basis.

Dietram Scheufele — a professor of science communication at the University of Wisconsin-Madison — and colleagues surveyed 1,600 members of the general public about their attitudes toward gene editing. The results revealed that 65 percent of respondents thought that germline editing was acceptable for therapeutic purposes.

When it came to enhancement, only 26 percent said that it was acceptable and 51 percent said that it was unacceptable. Interestingly, attitudes were linked to religious beliefs and the person’s level of knowledge of gene editing.

“Among those reporting low religious guidance,” explains Prof. Scheufele, “a large majority (75 percent) express at least some support for treatment applications, and a substantial proportion (45 percent) do so for enhancement applications.”

He adds, “By contrast, for those reporting a relatively high level of religious guidance in their daily lives, corresponding levels of support are markedly lower (50 percent express support for treatment 28 percent express support for enhancement).”

Among individuals with high levels of technical understanding of the process of gene editing, 76 percent showed at least some support of therapeutic gene editing, while 41 percent showed support for enhancement.

But how do the views of the general public align with those of genetics professionals? Well, Alyssa Armsby and professor of genetics Kelly E. Ormond — both of whom are from Stanford University in California — surveyed 500 members of 10 genetics societies across the globe to find out.

Armsby says that “there is a need for an ongoing international conversation about genome editing, but very little data on how people trained in genetics view the technology. As the ones who do the research and work with patients and families, they’re an important group of stakeholders.”

The results were presented yesterday at the American Society for Human Genetics (ASHG) annual conference, held in Orlando, FL.

In total, 31.9 percent of respondents were in favor of research into germline editing using viable embryos. This sentiment was more particularly pronounced in respondents under the age of 40, those with fewer than 10 years experience, and those who classed themselves as less religious.

The survey results also revealed that 77.8 percent of respondents supported the hypothetical use of germline gene editing for therapeutic purposes. For conditions arising during childhood or adolescence, 73.5 percent were in favor of using the technology, while 78.2 percent said that they supported germline editing in cases where a disease would be fatal in childhood.

On the subject of using gene editing for the purpose of enhancement, just 8.6 percent of genetics professionals spoke out in favor.

“I was most surprised, personally,” Prof. Ormond told Medical News Today, “by the fact that nearly [a third] of our study respondents were supportive of starting clinical research on germline genome editing already (doing the research and attempting a pregnancy without intent to move forward to a liveborn baby).”

This finding is in stark contrast to a policy statement that the ASHG published earlier this year, she added.

According to the statement — of which Prof. Ormand is one of the lead authors — germline gene editing throws up a list of ethical issues that need to be considered.

The possibility of introducing unwanted mutations or DNA damage is a definite risk, and unwanted side effects cannot be predicted or controlled at the moment.

The authors further explain:

“ Eugenics refers to both the selection of positive traits (positive eugenics) and the removal of diseases or traits viewed negatively (negative eugenics). Eugenics in either form is concerning because it could be used to reinforce prejudice and narrow definitions of normalcy in our societies.”

“This is particularly true when there is the potential for ‘enhancement’ that goes beyond the treatment of medical disorders,” they add.

While prenatal testing already allows parents to choose to abort fetuses carrying certain disease traits in many places across the globe, gene editing could create an expectation that parents should actively select the best traits for their children.

The authors take it even further by speculating how this may affect society as a whole. “Unequal access and cultural differences affecting uptake,” they say, “could create large differences in the relative incidence of a given condition by region, ethnic group, or socioeconomic status.”

“Genetic disease, once a universal common denominator, could instead become an artefact of class, geographic location, and culture,” they caution.

Therefore, the ASHG conclude that at present, it is unethical to perform germline gene editing that would lead to the birth of an individual. But research into the safety and efficacy of gene editing techniques, as well as into the effects of gene editing, should continue, providing such research adheres to local laws and policies.

In Europe, this is echoed by a panel of experts who urge the formation of a European Steering Committee to “assess the potential benefits and drawbacks of genome editing.”

They stress the need “to be proactive to prevent this technology from being hijacked by those with extremist views and to avoid misleading public expectation with overinflated promises.”

But is the public’s perception really so different from that of researchers on the frontline of scientific discovery?

The need

At the first International Summit on Human Gene Editing in December 2015, the organizing committee issued a statement about appropriate uses of the technology (see About the issue of making genetically modified children, it concluded that “it would be irresponsible to proceed with any clinical use … unless and until (i) the relevant safety and efficacy issues have been resolved … and (ii) there is broad societal consensus about the appropriateness of the proposed application”.

This should have been understood to mean that clinical uses of germline editing should not yet proceed anywhere in the world. Yet, subsequent events suggest that this statement was inadequate.

First, in China, biophysicist He Jiankui reportedly edited embryos to create at least two babies. Second, scientists who were apparently aware of this work did not take adequate measures to stop it. Third, there has been growing interest in proposals for genetic enhancement of humans 2 , 3 . Fourth, some commentators have interpreted subsequent statements as weakening the requirement for broad societal consensus 4 such statements include a 2017 report from the US National Academies of Sciences, Engineering, and Medicine 5 and a 2018 statement from the organizing committee following the Second International Summit on Human Genome Editing (see Finally, no mechanism was created in the ensuing years to ensure international dialogue about whether and, if so, when clinical germline editing might be appropriate.

A global moratorium and framework are therefore necessary to ensure proper consideration of the relevant issues surrounding clinical uses of germline editing.

Technical considerations. For germline editing to even be considered for a clinical application, its safety and efficacy must be sufficient — taking into account the unmet medical need, the risks and potential benefits and the existence of alternative approaches.

Although techniques have improved in the past several years, germline editing is not yet safe or effective enough to justify any use in the clinic. As was evident at the second summit, there is wide agreement in the scientific community that, for clinical germline editing, the risk of failing to make the desired change or of introducing unintended mutations (off-target effects) is still unacceptably high. Considerable research is being directed at this issue.

Scientific considerations. No clinical application of germline editing should be considered unless its long-term biological consequences are sufficiently understood — both for individuals and for the human species.

Among the vast array of possible genetic modifications, it is useful to distinguish between ‘genetic correction’ and ‘genetic enhancement’.

By genetic correction, we mean editing a rare mutation that has a high probability (penetrance) of causing a severe single-gene disease, with the aim of converting the mutation into the DNA sequence carried by most people. Assuming that it can be done without errors or off-target effects, genetic correction could have a predictable and beneficial effect.

Genetic enhancement, by contrast, encompasses much broader efforts to ‘improve’ individuals and the species. Possibilities range from attempting to modify the risk of a common disease by replacing particular genetic variants with alternative ones that occur in the human population, to incorporating new instructions into a person’s genome to enhance, say, their memory or muscles, or even to confer entirely new biological functions, such as the ability to see infrared light or break down certain toxins.

Understanding the effect of any proposed genetic enhancement will require extensive study — including of human population genetics and molecular physiology. Even so, substantial uncertainty would probably remain.

A human embryo at the eight-cell stage. Credit: Yorgos Nikas/SPL

Modifying disease risk by replacing genetic variants with alternative ones is fraught with challenges, because variants that decrease the risk of some diseases often increase the risk of others. A common variant in the gene SLC39A8, for instance, decreases a person’s risk of developing hypertension and Parkinson’s disease, but increases their risk of developing schizophrenia, Crohn’s disease and obesity 6 . Its influence on many other diseases — and its interactions with other genes and with the environment — remains unknown.

It will be much harder to predict the effects of completely new genetic instructions — let alone how multiple modifications will interact when they co-occur in future generations. Attempting to reshape the species on the basis of our current state of knowledge would be hubris.

The work of He illustrates this point. Seeking to decrease the children’s risk of acquiring AIDS if exposed to HIV later in life, He attempted to inactivate the gene CCR5, which encodes a receptor that HIV uses to enter cells. However, this change is not benign: it has been reported to substantially increase the risk of complications, and death, from certain other viral infections, including West Nile virus and influenza. It could have other consequences, too — both positive and negative (see Nature (2018) and ref. 7). As a societal solution to AIDS, disrupting CCR5 through clinical germline editing is ill-advised. Germline editing would not help individuals with the infection today, and it would require many decades of widespread use to make a dent in the epidemic. And, if an effective HIV vaccine is developed, the genetic enhancement would confer no benefit with respect to AIDS, yet still increase the risk of complications from other infections.

Medical considerations. Clinical application should be considered only if there is a sufficiently compelling reason. At the early stages of the new technologies, the bar should be set high.

Genetic enhancement of any sort would be unjustifiable at this time, given the scientific considerations already mentioned. The issue of genetic correction is more complex.

Some argue, especially in the popular press, that germline editing is urgently needed to stop children from being born with severe genetic diseases. But couples who know they are at risk of transmitting a severe disease-causing mutation already have safe ways to avoid doing so. They can use in vitro fertilization (IVF) in conjunction with preimplantation genetic testing (PGT), prenatal testing, sperm donors, egg donors, embryo donors or adoption. In particular, use of IVF followed by genetic screening of the embryos to ensure that only unaffected ones are transferred to the person’s uterus, guarantees that a couple will not have children with the genetic disease.

The real problem is that most children with severe genetic diseases are born to couples who did not know they were at risk. Routine access to preconception genetic screening could allow most at-risk couples to make use of current options, should they wish to do so. Better access to newborn screening is also needed, to ensure that babies with a genetic disease can immediately receive any available therapy.

What then is the role for genetic correction? Although IVF coupled to PGT can ensure that couples carrying a severe disease-causing mutation will not have an affected child, it doesn’t always yield a baby.

In most cases, the problem stems from limitations of the process, related to the number and quality of the eggs harvested, and the growth and implantation of the embryos produced. IVF itself does not always succeed the transfer of an embryo leads to a successful pregnancy in roughly 30% of cases in women under 35 and in less than 10% of cases in women over 40. PGT diminishes the number of embryos available for transfer, because some embryos are rejected as a result of the genetic testing results, and others fail to develop in vitro to a stage and quality that makes them suitable for testing.

In most cases, suitable embryos are available for transfer following PGT. However, when only a few are available to begin with, there might be no suitable ones after the test. Couples can repeat the process, and they might succeed on subsequent tries, but some might never obtain unaffected embryos.

It has been suggested that, if germline editing were highly efficient and safe, it might increase the proportion of couples that achieve pregnancies. However, continuing to improve the efficiency of the IVF and PGT processes might be a better, safer, cheaper and more widely applicable solution.

At present, it is hard to evaluate the case for using germline editing to improve the efficiency of IVF coupled to PGT. The extent to which PGT diminishes IVF efficiency as a function of the IVF protocol, age of the mother, number of eggs harvested, and proportion of affected embryos has not been extensively investigated. (We know of only one study addressing a few of these questions in a single setting 8 .) The efficiency of germline editing is also unclear, especially given the need to assess embryos for editing accuracy. Once these issues are clarified, the case can be weighed.

For a tiny fraction of couples, the situation is different. These couples can never be helped by IVF coupled to PGT alone, because 100% of their embryos will be affected. In these cases, one parent is homozygous for a dominant disease or both parents are homozygous for a recessive disease. Such instances are exceedingly rare, occurring for only a small minority of genetic diseases and largely in situations in which a disease allele is present at high frequency in a population.

These rare couples might represent the strongest case for considering clinical germline editing, because the technology would be their only way to conceive unaffected children who are biologically related to both parents. Societies will need to weigh the legitimate interests of such couples against other issues at stake.

Societal, ethical and moral considerations. Irrespective of all of the above, clinical germline editing should not proceed for any application without broad societal consensus on the appropriateness of altering a fundamental aspect of humanity for a particular purpose. Unless a wide range of voices are equitably engaged from the outset, efforts will lack legitimacy and might backfire.

The societal impacts of clinical germline editing could be considerable. Individuals with genetic differences or disabilities can experience stigmatization and discrimination. Parents could be put under powerful peer and marketing pressure to enhance their children. Children with edited DNA could be affected psychologically in detrimental ways. Many religious groups and others are likely to find the idea of redesigning the fundamental biology of humans morally troubling. Unequal access to the technology could increase inequality. Genetic enhancement could even divide humans into subspecies.

Moreover, the introduction of genetic modifications into future generations could have permanent and possibly harmful effects on the species. These mutations cannot be removed from the gene pool unless all carriers agree to forgo having children, or to use genetic procedures to ensure that they do not transmit the mutation to their children.

Ethics declarations

Ethics approval and consent to participate

All mouse work was approved by local Institutional Animal Care and Use Committees at the Medical College of Georgia (protocol #2019-0999 and #2019-1000) and Cornell University (protocol #2000-0122).

Consent for publication

Competing interests

DRL is a consultant and co-founder of Editas Medicine, Pairwise Plants, Beam Therapeutics, and Prime Medicine, companies that use genome-editing technologies. KH, JAW, and APK are employees of Synthego Corporation. CRL and SQT have filed a patent application on CHANGE-seq. SQT is a member of the scientific advisory board of Kromatid.