323 - CRISPR and the future of gene editing: scientific advances, genetic therapies, & more

KEYPOINTS

• 01

  Zhang came to gene editing from optogenetics and a need to target specific cells.

  Zhang, who finished his PhD in 2009 and started his lab at MIT and the Broad Institute in 2011, helped develop optogenetics—inserting light-sensitive genes into specific brain cells. That worked only in genetically modified mice, which took one to two years to engineer, motivating a faster, more general way to edit genomes.

• 02

  CRISPR began as a curiosity in bacterial DNA in the 1980s.

  Japanese researchers studying E. coli noticed unusual repeating sequences; the acronym CRISPR came in the early 2000s, and Francisco Mojica found that the spacer sequences between repeats matched viruses. Scientists initially focused on the repeats—the strongest signal—before realizing the spacers were an immune memory.

• 03

  How bacterial CRISPR works: a guide RNA directs a Cas enzyme to cut.

  After infection, bacteria store a snippet of viral DNA. On reinfection, the CRISPR region produces RNA carrying a roughly 20–30 letter guide that matches the virus; Cas proteins (numbered Cas1, Cas2… roughly by order of discovery) bind it, find the matching sequence next to a PAM motif, and cut. Viral genomes range from about 10,000 to 100,000+ letters.

• 04

  Earlier tools—meganucleases, zinc fingers, TALENs—were cumbersome.

  Each zinc finger recognizes three DNA letters, so six fingers recognize 18—enough for uniqueness because 4^18 exceeds the human genome's 3 billion letters. TALENs recognize single letters via 12–20 repeats. But both required engineering a new protein for each target, whereas CRISPR needs only a new guide RNA.

• 05

  Over 5,000 single-gene disease mutations make editing a powerful therapeutic idea.

  Since the Human Genome Project, researchers have identified more than 5,000 mutations directly causing disease, usually affecting small populations. In a dish, editing efficiency can approach 100%.

• 06

  Delivery is now the main bottleneck.

  Cas9 is about 1,300 amino acids and other Cas proteins around 1,000, too large for some vehicles. Ideally the system would be around 300 amino acids. The liver is robust—80–90% editing is achievable—while other tissues are harder. Solving delivery wouldn't solve cancer, Zhang notes, because cancers arise from many different mutations.

• 07

  Sickle cell disease is the first approved CRISPR therapy.

  Cells are edited outside the body and returned, reactivating fetal hemoglobin—an ex vivo approach that avoids the delivery problem but is costly and demanding.

• 08

  Base editing and AI-designed proteins are expanding the toolkit.

  Base editors change single letters without cutting both strands. DeepMind's AlphaFold 2 (2020–21) allows predicting structure from sequence, helping engineer proteins; Zhang's lab studies IscB, an ancestral Cas9 form only about 450 amino acids—roughly a third the size.

• 09

  Cas13 targets RNA and enabled diagnostics.

  Cas13, the RNA analog of Cas9, cleaves the target RNA and then indiscriminately cuts nearby RNA—a 'suicide' feature that makes a signal. Zhang's team used it for diagnostics during COVID.

• 10

  Germline editing for enhancement is far off and ethically fraught.

  Zhang discusses He Jiankui's 2018 editing of embryos for HIV resistance, done without proper consent and with little rationale since transmission could be prevented otherwise. Nobody knows how to make a child '30 IQ points smarter'; huge risks need huge payoffs.

• 11

  Gene therapy's long-promised revolution is finally becoming practical.

  Peter recalls that when the human genome was sequenced nearly 25 years ago, gene therapy was hailed as imminent, then disappointed. Zhang says the difference now is the ability to go to a precise address and fix it—something impossible 10–15 years ago. Still, therapies must be manufactured at clinical scale: his lab makes its own Cas9 and Cas13, and commercial supply is mostly laboratory-scale, while a human may need thousands of times the dose of a 10 g mouse.

• 12

  Efficiency multiplies, which limits benefit outside the liver.

  In the eye or brain, delivery may reach only some cells, and editing may work in only some of those; multiplied together, the fraction corrected can be small. That's one reason partial vision restoration in LCA trials was modest, and why age of treatment and neuroplasticity windows may also matter.