Rupert Beale

The business of science is intensely frustrating. Most experiments fail, most great ideas come to nothing, and most genuine discoveries turn out to be of modest importance. Years of effort can easily be wasted on what turns out to be a mirage. In biology, we usually fail for the dullest of reasons: a test wasn’t as specific as we thought, a wondrous result proved to be a simple mistake, the supposedly seminal paper on which we relied was bogus. Faced with an avalanche of failure, rare advances of limited importance are to be cherished. With CRISPR-Cas9, the gene-editing technique that has transformed molecular biology, many scientists were aiming to produce just this sort of solid result and instead discovered something world-changing.

In 1915, the English microbiologist Frederick Twort reported that some of his bacterial colonies had ‘glassy’ areas, where dead bacteria had burst open. He discovered that he could transmit the glassy appearance between colonies, and that the agent responsible could pass through porcelain filters, which indicated that it was much smaller than a bacterium. Two years later, the French-Canadian scientist Félix d’Herelle made a similar discovery while attempting to turn maple syrup into alcohol. After working out that these agents – viruses that d’Herelle called bacteriophages – could kill bacteria, he turned his attention to using them as a therapy for bacterial infection. He got as far as treating a case of dysentery in a chicken, but the therapy still hasn’t been effective in humans, and antibiotics subsequently made the approach seem redundant. It’s quite possible that antibiotic resistance will make bacteriophage treatment important once again, but for a long time bacteriophages were a scientific backwater.

In 1993, Francisco Mojica, working in a literal backwater, made the first in a series of curious discoveries while studying bacteria that live in Spanish swamps. He found that they contained repeated sequences of DNA that didn’t correspond to anything previously seen. The sequences occurred at regular intervals in the genome and read the same forwards and backwards: he called them ‘clustered, regularly interspersed, short palindromic repeats’, or CRISPR. DNA of similar structure had been discovered in several other bacterial species; Japanese researchers had spotted it in E.coli in 1987, for example. The sequences did not resemble one another, but they were like sequences from bacteriophages, which suggested that their function was to fight off these invading viruses. This was a very important advance but didn’t attract the attention it deserved until 2007, when microbiologists at Danisco – a once Danish dairy now owned by the American conglomerate Dow Dupont – provided the first experimental evidence that the CRISPR system enables bacteria to ‘remember’ viruses that infected them. Their research, which clearly had significance beyond industrial yoghurt production, was published in Science.

Immunological memory, the ability to respond more vigorously to an infection the second time round, had previously been thought to occur only in vertebrates. It’s the reason people usually only get chickenpox once, and that vaccination is effective. The way it works in bacteria is completely different from the way it works in humans and other animals. We rely on recognising the proteins from bacteria and viruses as different from our own; bacteria recognise the DNA of their bacteriophage aggressors. They can capture and copy a piece of the invading phage DNA. This copy is inserted into the bacterial genome as a CRISPR sequence. The bacteria then use this DNA as a template to allow them to recognise similar sequences in invading phages.

DNA is a very stable molecule used by all organisms (except some viruses) to store genetic information. Its structure – two complementary strands entwined in a double helix – allows it to replicate itself: the two strands separate, and each is used as a template to create a new version of the other, making it possible for a cell to divide into two identical daughter cells. DNA is also used as a template for producing RNA, a less stable molecule that can perform all sorts of functions within the cell. One of its important roles is as a messenger: it is the chemical intermediary between genes and the proteins they make. The DNA genetic template instructs the production of messenger RNA, which can be decrypted by a molecular machine called a ribosome to form a protein. This is the central dogma of molecular biology: DNA makes RNA makes protein. In the case of CRISPR sequences, however, the RNA does not encode a protein. Instead, since it has been copied from an invading bacteriophage’s DNA, it forms a perfectly complementary portion of RNA to any similar phage’s genome. The bacteria also produce an enzyme called Cas9 (short for ‘CRISPR-associated protein 9’) which can cleave DNA, but is inactive unless it is bound to a CRISPR-generated RNA molecule that has found a perfectly complementary strand of DNA. In this case it can break the backbone of the invading DNA, chopping it to pieces. The CRISPR-generated RNA effectively guides the Cas9 to the section of the bacteriophage’s DNA that it matches. This means we can use guide RNA and Cas9 as a precise way of cutting DNA at a particular spot.

Various methods of cutting DNA in a test tube have been available for decades, but CRISPR makes it easily programmable within mammalian cells. We can control the production of the guide RNAs, and since Cas9 isn’t as toxic as you might expect there’s remarkably little collateral damage. When we target a cellular gene with CRISPR, the gene keeps being cut, and the cell keeps repairing the cut. But eventually the cell will make an error, and accidentally delete a bit of DNA (usually a single base pair). That destroys the perfect complementarity required for the guide RNA to recognise the targeted gene, so the Cas9 stops chopping. It also alters the genetic code so that instead of the correct protein, the ribosome that decodes the message gets a signal to stop. If you successfully target the beginning of a gene, no coherent protein can be produced – the gene is ‘knocked out’.

I spent a full year of my PhD trying and failing to knock out a gene in a chicken cell line (at that point, 15 or so years ago, the state of the art). With CRISPR this would take a few weeks, and I could choose any cell type I liked. It’s often helpful to find out what happens if a gene is missing in order to understand that gene’s biological purpose. If patients without a particular gene succumb easily to viruses, for example, it’s a fair bet that the gene is involved in immune response. One of the things my laboratory is working on at the moment is the way certain genes that we’ve knocked out using CRISPR affect the replication of viruses such as influenza in human cell lines. Because the process is so much easier with CRISPR we can knock out several genes we know are involved in the same biological process, and make multiple, independently generated knock-outs of these genes. That means there’s much less chance of our coming to incorrect conclusions on the basis of just one or two cell lines. It also means it’s much quicker to work out which components of a known biological process are involved in viral replication, and which are dispensable. This would be helpful if, for example, we wanted to make a drug that prevented a particular virus from replicating properly.

The CRISPR-Cas9 knock-out technique is so efficient, it’s possible to disable all the genes in the human genome in the same experiment. If you set things up correctly, you can obtain a pool of cells with the Cas9 gene, each of which has a guide RNA that targets just one of the twenty thousand genes in the human genome. Within two weeks, you can have a vast ‘library’ of around ten billion cells, each with a gene knocked out. We need these huge libraries so that any given gene will have been disabled in many different ways; it improves the statistical power of the experiment. One application of this is to expose cancer cell lines to a chemotherapy drug. Most of the cells will die, but those that survive will be the ones that have lost genes essential for responding to this particular chemotherapeutic agent – very helpful information if you are trying to come up with ways to stop cancer cells becoming resistant to drugs.

It’s also possible to look at which genes are required for a cell to survive. Many of these essential genes have known functions: the ones involved in DNA replication, for example. Several research groups have analysed the genes required for the replication of a number of human cell lines. As well as the obvious ones, they discovered a large number of genes with completely unknown functions that are nonetheless essential for cell survival. There’s still a lot of cell biology we not only don’t understand, but don’t know we don’t understand. Techniques using CRISPR-Cas9 to disrupt genes in cells that we can study in the lab are still being refined, but are rapidly becoming routine. Techniques involving massive genome-wide libraries of CRISPR knock-outs are still tricky, time-consuming and expensive, but they are immensely powerful and hugely informative.

As well as knocking out entire genes, CRISPR can be repurposed to perform more sophisticated tasks. For example, guides can be designed to make two cuts in the same gene, chopping out a specific portion. This can be used to investigate the function of a particular part of a gene, and in principle could be used to get rid of a faulty part. You could then try to paste something else in: part of a different gene, or a synthetic correction – a patch – to correct the faulty gene. Alternatively, you can make a mutant Cas9 that lacks its cutting function. One ploy we have used is to bolt onto that version of Cas9 proteins that activate other genes. This means you can design guide RNA sequences that individually target each gene in the human genome for activation – the opposite of a knock-out. This enables you, for example, to find genes that protect against chemotherapy or a virus, or to discover the genes that when turned on cause the cell to commit suicide. Cell suicide is important to avoid cancer and combat infection (a virus can’t replicate in a dead cell).

With CRISPR-Cas9 techniques we can kill genes, switch them on and, if we are lucky, replace bits of one gene with another. It doesn’t stop there: the guidance system can be employed to perform almost any function that can be bolted onto a protein, and many research groups are exploring such possibilities. CRISPR has accelerated our progress in understanding fundamental biological mechanisms and will help us design new drug therapies. Will it be widely used as a treatment itself, or will it go the way of d’Herelle’s dream of phage therapy? It certainly offers many advantages over current gene therapies, but it doesn’t solve the big problem of gene delivery. If we are engineering cells in a petri dish, we can easily get genes into them using a viral delivery system. But our cells have evolved powerful mechanisms to resist viral infection, and the physical barriers to delivering genes to whole organs are considerable. It’s likely that CRISPR gene therapy will soon have a role to play in the treatment of disorders of white blood cells, for example: bone marrow can be removed from a patient, CRISPR-modified using specially designed viruses, and returned to the patient. My patients with genetic kidney diseases, however, are unlikely to be significantly helped by CRISPR in the near future, though it’s possible the new biology we discover could lead to new therapies.

For those of us who once spent months or years persuading even one gene to behave itself as instructed, CRISPR excites an eager, near childish joy. The first time my German postdoc and I got results back from genome-wide CRISPR screens, we stared at the results for a few moments to check we weren’t hallucinating, then burst out laughing and high-fived each other – highly uncharacteristic behaviour for both of us. CRISPR is a scientist’s chocolate factory, with pitfalls for the greedy. It isn’t just big news for scientists, it’s big business, and rival institutions are engaged in an unseemly squabble over the key patents – in the latest round, the claims of MIT and Harvard won out over those of UC Berkeley.

Jennifer Doudna, who works at Berkeley, made some of the seminal discoveries in the field and is one of a handful of people who could plausibly get a Nobel Prize for CRISPR. She has written a breezy popular science book about it, which explains the molecular details of CRISPR in an engaging way.[*] The final chapter looks at the prospects for using CRISPR to edit human embryos. The first possibility is that certain genetic diseases could be eliminated. Would this be safe? Who would benefit? Given the commercial interests at stake, is it possible we will create a genetic underclass? Should parents be allowed to edit their future children’s genomes? Whenever there is a breakthrough in genetic engineering, there are inevitably fears that it may be used irresponsibly in human reproduction. The nature of the arguments hasn’t been changed by CRISPR, but it’s more urgent that we address them. Doudna has thought about the issues, but it isn’t clear that she goes beyond paying lip service to some of the objections. It seems to be a glib remark from a student – ‘But what if we don’t edit the human germline?’ – that persuades her it’s all OK.

Enthusiasts for human germline editing got a fillip in August last year when a report in Nature claimed success in editing a mutant gene responsible for a serious heart disease in human embryos using CRISPR. This apparently wasn’t accompanied by any unintended effects on other genes. The report led worldwide news bulletins and there were triumphant front-page headlines. Too good to be true? Several leading scientists in the field think so, and have produced a polite rebuttal of the central claims. There’s no suggestion of fraud; it’s merely that less exciting interpretations of the published data seem more plausible: there may have been unintended deletion of genes, for example – an undesirable off-target effect. These doubts have received hardly any coverage. Most reporting on CRISPR oversells the technology; it’s an amazing research tool, but not a panacea. Doudna isn’t alone in reaching for references to Brave New World. She reckons that humans will be divided into different genetic castes a lot sooner than 2540. I’m not sure whether she thinks this desirable or merely inevitable; I’m sure it’s neither.

[*] A Crack in Creation: The New Power to Control Evolution by Jennifer Doudna and Samuel Sternberg (Bodley Head, 304 pp., £20, June 2017, 978 1 84792 381 3).