Stuart Newman, with whom I am working on a book (with Lenny Moss as the third author), drew my attention to this article from the Wall Street Journal by Preetika Rana and Lucy Kramer. It is an important article responding to the recent reports of successful germ-line editing from China (primarily the work of Dr. He Jiankui in Shenzhen—see, for instance, my earlier post). Rana and Kramer note the outrage generated by that experiment and the claims that the work was premature and ill-conceived. Then they note:
The proliferation of similar experiments on farm animals in recent years supports those concerns. Though rapid strides have been made to map genomes—the full set of genes for humans, animals, insects and plants—scientists have only begun to understand what the tens of thousands of individual genes do. Moreover, they are far from unraveling how those genes interact with each other.
I agree with these concerns and will discuss them in greater detail below. Since I have defended Dr. He against some of the criticisms directed against him, I should also note that I do find an ethical breach in his work but think it is one promoted by many of the enthusiasts of the Human Genome Project and subsequent genomics. That breach consists of appealing to fairy tale science in the context of medical intervention, in particular, the belief that single genes (or perhaps a very small number of genes) are competent in producing the vast majority of organismic traits.
The WSJ article details how attempts to modify animal traits through CRISP/Cas9-based gene editing led to unanticipated alterations in other traits. Attempting to enhance the quality of meat in pigs by deleting the MSTN gene led to leaner pigs but, in 20 percent of the offspring, an extra thoracic vertebra. Modifying the same gene in rabbits led to almost half the animals being born with enlarged tongues. In lambs, MSTN deletion led to birthing difficulties. The examples given in the article appear impressive enough though proponents of germ-line engineering can point out that almost all the examples implicate problems only when the MSTN gene is targeted and the the vast majority of genes of interest may well be immune to such problems.
But that is exactly what is unproven. Proponents of the widespread use of gene editing seem to be living in the very earliest years of genetics (which lasted from 1900 till at most 1910) when there was some reason to endorse what came to be known as the one gene-one character fallacy, that there is a one-to-one correspondence between genes and traits. By 1900 two phenomena were widely recognize: pleiotropy, which means that a single gene influences multiple traits; and epistasis, that multiple genes interact with each other in the development of a trait. Both phenomena point to the complexity of the relationship between genes and traits—I will probably have much to say on this on this blog (as it has been a major theme of my work). These complexities have long been known to classical geneticists but, in my opinion, are often ignored in gene editing work even though they are part of the rhetoric of genomic with its dominant metaphor of gene regulatory networks.
Consequently, when it comes to formulating policy about the permissibility of germ-line modification, we must go beyond the two criteria that have usually been invoked in these discussions: safety and accuracy. Safety relates to the question of whether the modification is harmful to health, particularly relevant to situations in which health and suffering are presumed to be of normative import, for instance, in vertebrates in experiment conducted in the United States. Accuracy refers to the ability to target the intended gene exactly, that is to target it without off-target effects. CRISPR-Cas9 is widely believed to have brought gene editing into a new era of unparalleled accuracy.
But, in my view, these two criteria are not sufficient: we also need what I call specificity of gene action, that modification of the intended gene results in an effect on the intended phenotype and only that intended phenotype. Specificity was introduced as a property of genes in Germany in the late 1920s in the context of the introduction of penetrance and expressivity—Manfred Laubichler and I reconstructed part of that story a long time ago (and I promise an entry on it later). I believe that all three are relevant to formulating policy on germ-line editing (in all organisms).
Here, I will restrict attention to the specificity of gene action. The MSTN cases noted earlier all violated this criterion. What the criterion does is address a standard objection to gene editing, whether it be somatic or germ-line: unintended consequences. Satisfying the criterion would require much work. Showing, experimentally, specificity of MSTN targeting in pigs (which I believe to be very unlikely) would not have demonstrated its specificity in rabbits. Moreover, specificity of any gene modification would require sampling of all possible genetic and developmental backgrounds. It is a tall order and, thus, I believe will be resisted by those who advocate germ-line modification.
Perhaps there is an argument to be made that when organisms are being grown purely for human use safety is enough and specificity irrelevant. For instance, the pigs with extra vertebrae mentioned earlier apparently remained safe to eat. But this argument has no relevance to human germ-line modification. Returning to Dr. He’s purported experiments, on the one hand I continue to be sympathetic to him because requirements such as specificity have not been imposed either by the international community of biologists through international guidelines (and, as far as I have been able to gather, not by Chinese institutions either). But, nevertheless, it is not clear that he had addressed safety issues sufficiently and, moreover, in not broaching specificity he has violated a minimal requirement of using the best science available. But he is hardly alone in the gene editing community.