Peter Attia MD323 - CRISPR and the future of gene editing: scientific advances, genetic therapies, & more
CHAPTERS
- 0:00 – 10:33
Feng Zhang's optogenetics work and the pivot to gene editing
Feng Zhang explains the genome-sequencing context for fixing disease mutations. Peter Attia then walks through Zhang's PhD with Karl Deisseroth on optogenetics. Zhang says precise targeting of cell types in the genome was a bottleneck that pushed him toward making gene editing easier.
- •Sequencing genomes let researchers identify more than 5,000 mutations that directly cause genetic disease
- •Optogenetics uses a light-sensing gene from green algae, delivered by engineered virus, to control brain cells
- •Targeting specific cell types required inserting genes at precise genome locations, which was hard
- •Transgenic mouse tricks took a year or two and could not be used in humans
- 10:33 – 21:49
The discovery of CRISPR repeats in bacterial DNA
Attia and Zhang trace CRISPR back to the 1980s observation of clustered repeats in bacterial genomes. They cover the acronym, Francisco Mojica's insight that the spacers match viral sequences, and the identification of Cas genes next to the repeats.
- •Japanese researchers found regularly spaced repeats in E. coli in the 1980s
- •CRISPR stands for Clustered Regularly Interspaced Short Palindromic Repeat, a name coined in the early 2000s
- •Mojica found that the non-repeating spacers matched viral sequences
- •Genes next to the repeats encode nucleases, called Cas proteins
- •Cas proteins pair with CRISPR RNA, which acts as a guide to find viral targets
- 21:49 – 30:06
How CRISPR defends bacteria: adaptive immunity, Cas9 and the PAM
Zhang describes how bacteria acquire and use viral memory. On first infection a few survivors store a ~30-letter viral snippet. On reinfection the guide RNA directs Cas9 to cut viral DNA, and the PAM sequence prevents self-targeting.
- •CRISPR is an adaptive immune system, though bacteria have other defenses like restriction endonucleases
- •Survivors of a first infection insert a ~30-letter viral snippet into the repeat array
- •On reinfection, many Cas9-guide RNA copies search in parallel and cut the virus within minutes
- •The PAM sequence is present in viral DNA but absent from the bacterial CRISPR array, so Cas9 does not cut the bacterium's own genome
- 30:06 – 39:08
Zinc fingers and TALENs: gene editing before CRISPR
Zhang recounts how he first encountered CRISPR in 2011 and what the existing editing tools were. Zinc finger nucleases and TALENs recognized DNA with proteins, and both were cumbersome to engineer for each new target.
- •Zhang read early Cas9 papers and saw an RNA-guided DNA-cutting enzyme
- •Zinc fingers recognize three letters each and are tethered into arrays, but are very hard to engineer
- •TALENs come from a rice pathogen and have a programmable one-letter code, but take weeks to months to build
- •Repetitive sequences are prone to recombination and hard to assemble
- 39:08 – 47:23
Why CRISPR is a smartphone-like step change
Attia reflects on the gap between sequencing the genome and being able to edit it. Zhang explains his shift to CRISPR in his new lab and compares it to a smartphone where the guide RNA is the software.
- •Knowing a mutation's location did not mean it could be fixed a decade after the genome was sequenced
- •Zhang started his lab focused on gene editing and worked on both TALENs and CRISPR
- •Zinc fingers and TALENs required new hardware for each gene
- •CRISPR guide RNAs are easy to synthesize, democratizing editing
- 47:23 – 57:24
Cutting, repairing and inserting DNA, and what diseases are targets
Zhang explains how a Cas9 cut triggers either error-prone repair that inactivates a gene or template-based homology-directed repair. Inserting large genes remains inefficient. Examples include liver amyloidosis and Huntington's disease.
- •Two repair pathways follow a cut: gluing that can inactivate a gene, or HDR using a template
- •Inserting a large gene precisely is still only single-digit percent efficient
- •Inactivating toxic gene products can treat liver amyloidosis
- •Huntington's requires precise three-letter deletions in a long repeat plus a brain-delivery method
- 57:24 – 1:06:51
Delivery challenges and sickle cell therapy
Attia recaps the Cas9 and guide RNA components and asks how to deliver them. Zhang describes the ex vivo approach for sickle cell disease, where harvested bone marrow stem cells receive Cas9 mRNA and guide RNA. The approved therapy edits a different gene to turn on fetal hemoglobin.
- •Cas9 can be introduced as DNA, mRNA or protein
- •Sickle cell stem cells are harvested, treated with Cas9 mRNA and guide RNA, and can reach near 100% editing in a dish
- •The approved therapy cuts a different gene to turn on fetal hemoglobin and needs no template repair
- •Base editing, developed by David Liu, uses a deaminase to change a single base without cutting
- 1:06:51 – 1:13:10
Payload versus delivery, cancer, and AI protein prediction
Zhang says editing payloads have advanced far, while delivery to the right cells is the bottleneck. They discuss why cancer is hard to cure by correcting mutations, and how AI such as AlphaFold2 predicts protein structure.
- •Payloads include Cas9, base editing, prime editing, recombinases, epigenetic editors, mRNA and siRNA
- •Cancer has many mutations and incomplete correction lets cells keep dividing
- •Gene editing is used to engineer immune cells for immunotherapy
- •AlphaFold2 learned from experimentally determined structures to predict protein shapes
- 1:13:10 – 1:17:43
CRISPR transforming transgenic mouse research
Attia and Zhang discuss how CRISPR speeds up building animal models. Injecting Cas9 and guide RNA into a single-cell embryo cuts a year or more of breeding to a few months, accelerating biomedical research.
- •Traditional transgenic mice used modified stem cells and produced mosaic animals that needed breeding
- •CRISPR can edit a single-cell embryo directly, with a 21-day gestation
- •Time to get a mouse drops from about two years to two or three months
- •Zhang is not worried about students missing the slower process and expects AI to help analyze growing data
- 1:17:43 – 1:27:46
Cas13, RNA targeting, diagnostics and smaller Cas proteins
Zhang explains Cas13 as the RNA-targeting analog of Cas9, with a collateral cleavage function used for COVID diagnostics. They then discuss the size problem of Cas9 for delivery and the search for compact systems like iscB.
- •Cas13 recognizes RNA and cleaves other RNA in the cell, shutting down the infected bacterium
- •This collateral activity enabled a paper-strip, point-of-care coronavirus test
- •Cas9 is about 1,300 amino acids and barely fits in viral vectors
- •iscB, the ancestral form of Cas9, is about 450 amino acids but needs a guide RNA three times larger
- 1:27:46 – 1:40:59
Therapies in the liver and eye, and the cost of gene editing
Zhang describes in vivo delivery: lipid nanoparticles to the liver and viral vectors to the eye. They cover Luxturna, the LCA10 editing trial, PCSK9 targeting, and why costs are high and likely to fall.
- •Lipid nanoparticles reach the liver, where 80–90% editing is possible
- •Luxturna puts a gene into the eye, while the LCA10 approach uses Cas9 to knock out a mutant gene
- •The LCA10 trial showed some improvement, but less than hoped, partly because the retina does not regenerate
- •PCSK9 inactivation could treat cardiovascular disease
- •Cost is driven by early-stage manufacturing, with scale expected to reduce it
- 1:40:59 – 1:50:38
Germline editing, the CCR5 embryo case and regulation
Attia raises the controversial 2018 embryo editing case in China. Zhang explains the CCR5 edit, the mosaic outcome, and the backlash because there was no medical need. He describes the resulting international working groups and the debate over germline editing.
- •Embryos were edited at CCR5 and the resulting girls were mosaic
- •The backlash arose because the edit was not medically justified
- •Multinational working groups were formed and US law prevents germline modification
- •Scientists agree the technology needs more validation of efficiency and specificity
- •The debate pits relieving suffering against the slippery slope toward designer babies
- 1:50:38 – 1:58:26
Ethical lines: disease treatment versus enhancement
Zhang gives his view that clear medical benefit can justify germline editing, with alternatives like pre-implantation testing considered. Both discuss why enhancing traits like intelligence is far harder than people think, given polygenic biology, environment and trade-offs, and why diversity matters.
- •Zhang finds clear medical benefit for inborn errors easier to agree on, while noting alternatives like pre-implantation genetic testing
- •Traits like intelligence are polygenic and shaped by the environment
- •Biology has compensation, so edits may carry trade-offs like cancer risk
- •Somatic treatments or drugs may offer similar outcomes with less permanence
- •Zhang says genetic diversity makes the human race resilient
- 1:58:26 – 2:13:15
Zhang's early life, education and path to gene editing
Zhang recounts moving from China to Iowa at age 11, a Saturday molecular biology class, Jurassic Park, and a high-school gene therapy lab where he saw glowing cells. He also discusses his Harvard chemistry and physics degree, choosing Stanford, and the Harvard Society of Fellows.
- •His mother moved to the US partly for its hands-on education
- •Teacher Ed Pelkington's class and a hospital gene therapy lab inspired him
- •He studied chemistry and physics because those foundations change less than biology
- •The Harvard Society of Fellows gave him freedom to start exploring gene editing
- •He credits teachers and the American merit-based education system
- 2:13:15 – 2:18:46
Optimism about science and the future of STEM
Attia asks whether Zhang remains optimistic despite concerns about meritocracy and public trust in science. Zhang says he is optimistic, pointing to accelerating data, technology, AI and robotics, and says more must be done to nurture curious young people.
- •Zhang says optimism is the only way to be
- •Sequencing, mass spectrometry, gene editing and microscopy are generating richer data faster
- •AI and robotic automation could create closed-loop experimentation
- •The education system should support curious kids with opportunities like those he had
