God’s Imperfect Red Pencil


From the perspective of noise, CRISPR systems are modern biology’s closest approximation to Beyonce’s Lemonade or Game of Thrones: buzz-worthy, trending, think-pieced-to-paralysis. CRISPR invites discussion from every possible angle – ethical, philosophical, economic, scientific -- a technology that seems to have both power and charisma.

The way that ‘CRISPR’ rolls off the tongue is evocative of its functional, scientific capacity. CRISPR is unquestionably the most powerful gene editing technology the world has yet seen, although folks like to debate more meaningless modifiers: is CRISPR super powerful, or just pretty powerful, and is this science fiction? Simply put, CRISPR is powerful.

CRISPR systems have a variety of useful applications, including new ways of imaging cells and changing levels of gene expression. Yet its most studied function thus far, and the way it has been introduced through the popular media, is as a tool for genome editing. CRISPR systems can change the sequence of letters that spell out our genomes and thus have the potential to alter almost any aspect of how any life form functions.

We have been able to edit genomes before, but CRISPR makes it unprecedentedly easy. Carl Zimmer of the New York Times has called it “Microsoft Word for gene editing” although a more stylish group might prefer “God’s red pencil.”

If CRISPR is a word processor, it is a remarkably bad one. Let’s say you are drafting an email to your parents, explaining your newest tattoo. You have written: “the spider on my face looks pretty great.” But you are reading it over and would like to be more descriptive: “The spider on my face looks pretty hairy.” With a CRISPR word processor, a number of problems might arise.

The most common and thorny problems are off-target effects. In addition to changing that sentence, you might end up replacing “great” with “hairy” at other places in your email. You now find your email saying things like, “You have always been hairy at respecting my decisions” or “I’m so glad I met John on Tinder – he is just so hairy!”

Also, some locations are simply difficult to edit. Maybe you want to take out the part that the spider is “on my face,” but you simply aren’t allowed to make that change with the CRISPR word processor. In that particular case, it’s just as well, since it will be hard to hide your face tattoo when you go home in December.

Overall, CRISPR works quite well, but there is distance between its reality and the image of it being as easy and precise as Microsoft Word. This is the same distance that exists between our idea of DNA as the genetic code of life and the reality of how DNA exists inside of a cell.

We can indeed think of DNA as a sequence of letters, but those letters are not just abstract information, they are also physical molecules. As such, DNA can interact with other molecules, most notably proteins. A strand of DNA is not simply loose spaghetti, but is often organized in nucleosomes, places where the DNA strand wraps twice around a core group of eight proteins called histones, and then continues on its way.

As we always find with biology, this process is messy and dynamic. Somes stretches of DNA are wrapped into nucleosomes and interspersed among these are regions of free, un-bound DNA. Like shirt collars, some nucleosomes are wrapped tightly and others more loosely and flirtatiously. Histones can also be modified: a tight collar gets unbuttoned, a loose one cinches up. Finally, nucleosomes can slide up and down a stretch of DNA, so it is hard to say whether a given region is nucleosome-bound or not, as this may change over time.

Two recent papers from UCSF labs, both published in ELife over the past two months, are addressing the reality of DNA organization in nucleosomes as it pertains to CRISPR gene editing. They show that some of the imperfections in CRISPR may be due to the presence of nucleosomes, but also suggest how CRISPR gene editing has been largely successful despite this limitation.

The first paper, published on March 17th, is a collaboration between the labs of Jonathan Weissman (UCSF) and Robert Tjian (Berkeley), while the second, published on April 28th, is a collaboration between the labs of Geeta Narlikar and Wendell Lim (both UCSF).

The papers make different points, but both consider the nucleosome as an obstacle for CRISPR gene editing. The actual molecular editing tool of CRISPR, most closely equivalent to the cursor in a word processing program, has two parts: a protein called Cas9 and a guide RNA. The guide RNA provides specificity, or is supposed to. It finds a matching sequence within the genome and binds to it, positioning Cas9 to make an incision at that spot, which is now opened up for making a number of changes.

But if the Cas9/RNA tool can’t access the DNA, because the DNA is wrapped up in nucleosomes, no edit will happen. On the flip side, off-target effects may be much more likely to happen in areas of open DNA. It is not just the sequence of the DNA that determines CRISPR efficacy, but the dynamic pattern of how that DNA is wrapped up with protein.

The first paper, led by Max Horlbeck from the Weissman lab and Lea Witkowsky from the Tjian lab, suggests that nucleosomes act as effective barriers for Cas9 access. They screened many possible editing sites within a human cell line, and show that effective CRISPR activity anti-correlates with the level of nucleosome present at any given point. They also show in vitro that Cas9 was unable to cut a DNA sequence bound within a nucleosome.

The second paper, led by Ricardo Almeida from the Lim lab and R. Stefan Isaac from the Narlikar lab, fleshes out the in vitro picture. They show that the ability of Cas9 to cut nucleosome-bound DNA depends on how tight that collar is, which often depends on the DNA sequence. The DNA sequence used in the first paper, the one that showed nucleosomes to block Cas9, was one that bound nucleosomes very tightly, thus preventing Cas9 access. However, the second paper showed that other, naturally-occurring DNA sequences, ones that bind DNA more loosely, in fact allow a high level of Cas9 access.

This is partly due to a phenomenon known as nucleosome breathing. This refers to the dynamic nature of the interaction between the DNA and protein, which is not fixed, but can oscillate over time: tightening, loosening, tightening, loosening. Cas9 may be able to enter at the ebbing moment and make its edit.

Both papers also showed that Cas9 could access and cut DNA when they included histone modifying proteins in their experiments. There are many of these proteins, some that slide the nucleosome along the DNA, some that get rid of histone binding entirely. They make the picture more fluid, more dynamic, providing more gaps and opportunistic moments for Cas9 to cut DNA.

Despite this complicated picture, some regions of DNA are simply more difficult for CRISPR systems to edit, often because of the presence of nucleosomes. Fortunately, we have decent knowledge of which DNA regions have affinity for nucleosomes, and thus can take these into account when designing CRISPR experiments.

The CRISPR biologist must be clever, designing sequences carefully to get past the histones to successfully modify DNA. Yet this is not to say that every obstacle presented has a neat solution. Off-target effects still loom as a large unsolved problem in CRISPR applications. Whether this is a fundamental limitation, or just a matter of de-bugging the word processor, is not yet known.