268 ‒ Genetics: testing, therapy, editing, association with disease risk, autism, and more
- Guest
- Wendy Chung, MD, PhD
- Read time
- ~5 min
- Uploaded
- 28 Aug 2023
- Duration
- 2:35:33
- Views
- 50.3K
KEYPOINTS
- 01
PKU is the model for newborn screening: catch it early and a disabling disease becomes manageable.
Wendy Chung, a clinical and molecular geneticist who is leaving Columbia to chair pediatrics at Boston Children's Hospital, started her career on phenylketonuria. It's a recessive disease: both parents are carriers and the child can't clear phenylalanine. Untreated, toxic byproducts damage the brain. A carefully titrated low-phenylalanine diet started in infancy prevents intellectual disability, and it's needed for life, especially during pregnancy. Through research she has found teenagers who were missed at birth and now have irreversible damage. Screening has moved from bacterial growth tests on heel-prick blood spots in the 1960s to tandem mass spectrometry. Her program adds DNA as a second, independent check, since neither the genetic sequence nor a single blood level perfectly predicts who needs treatment.
- 02
Gene discovery went from years to hours, and still only about a third of genes are linked to disease.
Chung's first cloned gene took eight years. Her most recent took hours. The cost of a genome fell from about $1 billion around 2000 to about $1,000 today. Humans turned out to have about 20,000 genes, not the predicted 100,000; complexity comes from each gene making many versions of its protein. Roughly 7,000 genes now have a known disease association. A standard genome reads short fragments of about 100 bases and maps them onto a reference. Some regions are never placed, and most clinical interpretation still focuses on coding regions. For a healthy middle-aged person the result is usually 'we don't see much here,' since they've already survived the most devastating variants.
- 03
Mutations build up over a lifetime, including in sperm and eggs.
Peter learned from family sequencing that he is mosaic for a variant one of his children inherited in full. Chung explains that new mutations arise with each cell division. In sperm they accumulate roughly linearly with paternal age, and some 'selfish' sperm mutations even help those cells divide. Maternal age raises chromosome errors such as Down syndrome more sharply from the mid-30s, because the machinery that separates chromosomes in eggs must work for decades after it starts in fetal life.
- 04
About 20% of cancer has a genetic component, and blood tests read what you were born with.
Chung estimates that about 5% of cancers come from powerful single genes with very high lifetime risk, another roughly 5% from moderate-risk genes that raise risk two- to threefold, and perhaps about 10% from polygenic risk scores summing hundreds to thousands of variants. Polygenic scores aren't yet used routinely in clinics. Cheek swabs and blood show inherited risk, not changes inside a breast. Liquid biopsies that look for tumor DNA in blood are, in her view, not ready for population screening; they're mainly useful for monitoring recurrence or high-risk people.
- 05
Consumer DNA tests are 'recreational' and can mislabel samples.
23andMe-style tests read selected single-letter variants (SNPs) to report ancestry and traits like earwax type or lactose intolerance. They don't read your whole genome and aren't built to answer medical questions, though they do report APOE status. Peter has seen several APOE results that didn't match later clinical sequencing. Chung notes mix-ups as simple as a flipped sample plate. Before anything irreversible, such as a mastectomy or a big insurance purchase, confirm with a clinical test.
- 06
The GUARDIAN study screens newborns' genomes for about 250 treatable conditions.
It began in New York City in September 2022 after her spinal muscular atrophy screening program, started in 2016. SMA had been the leading genetic cause of death under age 2. Screening found babies before symptoms, while the treatment window was still open. GUARDIAN had screened just over 3,000 babies at recording and aims for at least 100,000. It costs about $1,000 per baby including interpretation and counseling, with a goal of about $100. Many conditions found don't need gene therapy at all: Wilson disease, which can end in liver transplant, is treated with inexpensive zinc that crowds out copper.
- 07
Gene therapy delivery is still the hard part.
Jesse Gelsinger died around 2000 at Penn in a trial for ornithine transcarbamylase deficiency, after an overwhelming immune and liver reaction to the adenovirus used to carry the gene. Vector reactions have caused other deaths since. Doses are tricky: too high provokes immunity, too low wastes what may be the only chance, since the body then neutralizes a repeat dose. Few gene therapies are approved (SMA is the standout) and they cost about $1–3 million. The CRISPR-Cas9 system's weak point is that it cuts both DNA strands, risking edits in the wrong places. Prime editing nicks only one strand, which gives more accurate repairs. Tay-Sachs needs a missing enzyme added back to the brain, where even 20–50% of normal may be enough. Fragile X needs an oversized DNA repeat edited down.
- 08
Autism is highly heritable but spread across many genes.
Its heritability is about 0.8–0.9, compared with roughly 0.5–0.6 for bipolar disorder and schizophrenia, about 0.3 for major depression and about 0.5 for obesity. No single gene accounts for more than 1% of cases, and new mutations not inherited from either parent aren't counted in heritability. Prevalence is about 2%, a rise that reflects changing diagnostic criteria, better recognition (especially in underserved groups) and access to services. Chung runs the large SPARK autism study, which found a teenager whose autism stemmed from undiagnosed PKU. Most of what's known comes from people under 20, and some subtypes may be degenerative in adulthood.
EXAMPLES
- 01
Peter's PCSK9 mea culpa.
He remembers reading Helen Hobbs's papers on people with overactive and underactive PCSK9 about 15 years ago and predicting the idea wouldn't pan out. The drug class proved him wrong.
- 02
The aspartame warning label.
People with PKU follow a tightly controlled diet. Labels matter most for pregnant women, whose fetus can be harmed without any warning symptoms.
- 03
Weighing the price tag.
Peter contrasts about $1 million in chemotherapy for a year of life at the end of life with a $1 million infant gene therapy that could give 80–90 years.
- 04
The CRISPR babies case.
The edit of CCR5 to prevent HIV in IVF babies was a poor choice of use case, since sperm washing already prevents transmission. The scientific community rallied against germline editing.
OTHER RELEVANT
- 01
Lp(a) and FH.
High Lp(a) affects roughly 1 in 10 people. FH, defined by LDL cholesterol above 190 mg/dL, involves thousands of genes, so gene therapies would more likely mimic drugs, for example by switching off PCSK9.
- 02
Genetic cardiomyopathies.
They affect about 1 in 500 people. Early animal data for genetic therapies look promising, though Chung is wary of tinkering with the heart.
- 03
Where the line sits.
Current consensus is to treat the body, not edit the germline. Enhancements are 'a trigger word for we shouldn't go there.'
- 04
Chung's 2040 forecast.
She expects diagnosis to become trivial with cheap sequencing and AI. Therapies will take longer to reach people.