Much Ado About CRISPR Germline Editing

Much ado has been made about the potential for CRISPR technology to radically transform life in domains from agriculture to medicine. No doubt this gene-editing tool will improve crop yields and reduce reliance on harmful pesticides. It will also provide gene therapies for many crippling diseases. For example, people with sickle cell anemia can have their stem cells edited to remove the mutated HBB gene, then have their bone marrow irradiated and repopulated with the repaired cells, essentially curing their sickle cell anemia. This technique can be used for several other blood diseases as well, including AIDS.

This last type of molecular editing is what Chinese biophysicist He Jiankui performed on embryos, resulting in a sentence for He of three years in prison. His actions were deemed medically reckless, especially since there already existed established ways to prevent babies from contracting HIV from their mothers both in utero and during birth.

Although it is relatively uncontroversial to edit somatic cells—that is, cells that are within the body but whose DNA is not passed onto future generations—his team edited embryos, a much more fraught endeavor. That’s because the edited version of the gene would appear not only in every cell in the girls to be born but also in their offspring. This intergenerational form of gene editing has been highly controversial. Indeed, a National Academy of Sciences, Engineering, and Medicine panel has called for a moratorium on germline gene editing using CRISPR or related technologies.

Yet, at the same time, this application of CRISPR inspires the most sci-fi, utopian visions: visions of a world where sickle cell anemia is gone once and for all, along with Huntington’s disease, Tay-Sachs disease, and most other Mendelian disorders. A world in which schizophrenia and hypertension become rare, and perhaps we become a bit smarter, taller, and faster. Jennifer Doudna, co-winner of the 2021 Nobel Prize for Chemistry for her role in developing CRISPR, supports a pause but does not think that human germline editing should ultimately be banned.

However, the reality of reproductive genetics is that genetic editing of embryos, ova, and sperm (or their progenitor cells) is not feasible for most conditions of interest, and may not even be necessary. So, let’s stop fretting so much about regulating genetically modified humans and worry more about regulating genetically predicted humans that are being born today.

To illustrate why CRISPR is not necessary for a couple to have a healthy biological baby, let’s walk through various scenarios. We can start with a disease that is a single-gene disease—the ideal target for editing—and where the disease allele is dominant. In the case of Huntington’s disease, if one or both parents have the affected gene, then preimplantation genetic diagnosis (PGD) is available to screen embryos before implantation. Assuming only one parent is affected, then half the embryos should be healthy. Even if both parents are affected, a quarter of the embryos will still be Huntington’s-free. So, parents can implant a biologically related embryo without the disease allele and not have to resort to risky gene editing.

Tay-Sachs disease, by contrast, is a recessive disease that occurs most frequently among the Ashkenazi Jewish population. In this case, genetic screening of the parents to be certain that at least one parent is homozygous for the healthy allele is sufficient. (Indeed, professional matchmakers in Israel have made use of genetic information in their assignments.) If, however, both parents are carriers of the risk allele, then PGD can ensure that there are healthy embryos.

For single-gene diseases, gene editing of an embryo becomes the only option for obtaining an unaffected, biologically related offspring only when one parent is homozygous for the disease allele and that allele is dominant, or when both parents are homozygous for the risk allele when it is recessive. Such occasions are rare in the population. The Huntington’s disease gene appears in 1 in every 5,000 alleles in Western populations.¹ Tay-Sachs disease is fatal in early childhood, so only carriers (i.e., those with one good copy to pass on) reach reproductive age. Cystic fibrosis, another recessive, single-gene disease, has a higher allele frequency, with a prevalence in the white US population of about 1 in 3,000 births, so that the chances that two sufferers would mate by chance are markedly greater—1 in 9 million.²

Even so, in cases where we know the single genetic sequence that CRISPR would snip out and replace, genetic screening of parents and embryos is much safer and effective. Germline editing is simply not necessary. When we shift to polygenic diseases and traits, we move to a realm where CRISPR is not just less feasible but also virtually impossible.

We have long known that conditions like schizophrenia or cardiovascular disease are caused by more than one gene. Just how many more polygenic diseases there are has come as a huge surprise. When the Human Genome Project (HGP) completed the draft sequence of the so-called book of life in 2003,³ many scientists thought it would be just a few short years before the handful of genes that cause hypertension, the dozen or so involved in cardiovascular disease, and the score that confer risk for major depression were found. But the genetic architecture of these and most other diseases turned out to be highly polygenic. Rather than two dozen or even a hundred places in the genome that confer increased risk for most diseases, it is thousands of small tweaks that, together, raise or lower someone’s height or risk for dementia. Effects turned out to be so small and so widely distributed that Stanford population geneticist Jonathan Pritchard offered a theory he called the “omnigenic model” in which basically every gene is involved in most outcomes.⁴

We can’t really expect CRISPR to refashion life by a thousand cuts. On that scale of gene editing, off-target effects are sure to occur with unknown sequelae. Moreover, as the omnigenic model suggests, genes have multiple effects. So, even correctly “fixing” a gene (or a few hundred genes) in the hopes of reducing the risk of autism, for example, may lead to increased risk for some other disease. Such a tightly bound network of genes and their effects might have been expected when the human genome was first sequenced and we discovered that we had about 20,000 genes, but we now understand that genes perform multiple roles in different tissues and at different times in development. Thus, tweaking them to one end would send ripples through the network.

Finally, the thousands of markers we detect in genome-wide association studies aren’t  the causal variants, and changing them probably would have no effect on the disease we are trying to cure. A causal variant is a locus of genetic variation that has biological consequences.  When a locus somewhere in the genome confers risk through a genome-wide association study, all we know is that some DNA in that general area matters for the outcome. It’s like having the genetic zip code but not the street address. That’s because DNA travels in chunks, called linkage disequilibrium blocks. Nature is not a great shuffler of the genetic deck of cards, so many cards (i.e., nucleotides) travel together. When we discover a locus associated with a particular trait, scientists zero in on the causal site by performing analysis called fine mapping. Sometimes fine mapping involves collecting data from different ancestral groups that have different chunkings of DNA in order to narrow the signal through analysis of where chunks in different populations overlap and where they don’t. But more often, it involves carefully manipulating (i.e., editing) the DNA around the signal. This kind of experiment can be done in fruit flies, but not in humans, for obvious ethical reasons.

We should neither be fretting nor holding our collective breath for CRISPR to make us all taller, smarter, and healthier. However, while those thousands of markers that predict most traits and diseases aren’t of great use to genetic engineers, they could form the basis of antenatal genetic prediction therapies, since one does not need to know causal variants in order to predict which embryo will likely grow up to be taller, stronger, and fitter.

This form of embryo selection based on polygenic prediction is what we, as a society, should be discussing and, potentially, regulating. Preimplantation genetic diagnosis has been around for quite some time for single-gene diseases. But, in 2020, the first baby selected based on its polygenic index for a given trait was born. Many more are surely in utero now. If we leave the domain of reproductive medicine unregulated, as it currently is, we risk a brave new world dawning where we have, as depicted in the 1997 sci-fi film Gattaca, two populations: an “inferior race” born the old-fashioned way, and a new breed of humans selected for specific traits that their parents sought to maximize—along with whatever “off-label” effects may come along for the ride.

This is all to say that, for the foreseeable future, genetically modified humans are going to be a much less widespread phenomenon than are genetically selected humans. With this in mind, let’s have less ado about CRISPR germline editing and more discussion about antenatal genetic selection.

References

1. Kay C, Collins JA, Miedzybrodzka Z, et al. Huntington disease reduced penetrance alleles occur at high frequency in the general population. Neurology. 2016;87(3):282-288. doi:10.1212/WNL.0000000000002858.
2. Cystic fibrosis. Medline Plus website. https://medlineplus.gov/genetics/condition/cystic-fibrosis/#frequency. Accessed March 11, 2022.
3. Venter JC. A part of the human genome sequence. Science.2003;299(5610):1183-1184. doi:10.1126/science.299.5610.1183.
4. Boyle EA, Li YI, Pritchard JK. An expanded view of complex traits: from polygenic to omnigenic. Cell. 2017;169(7):1177-1186. doi:10.1016/j.cell.2017.05.038.

About the Author