TED gets epigenetics wrong, but the juggernaut rolls on

I was hoping that the Epigenetics Tsunami would have abated a bit by now, but it’s still washing over the public. And the new video below, produced by TED-Ed, doesn’t help, for it distorts what we know about epigenetics. The organization should be named “TED-anti-ED”:

Just a few of the erroneous claims that you’ll hear in the video:

  • All of differential regulation of genes during development (how an original undifferentiated cell turns into an organism having diverse tissues) is caused by epigenetic markers on those genes or on the histone proteins that are the scaffolds for genes. That’s not correct. The differential regulation occurs through the differential expression of transcription factors and small RNA molecules that activate or inactivate genes, with the heritability across mitotic generations caused by positive feedback. Any associations between “epigenetic markers” and gene expression are probably due to the modification of genes after they’re already been expressed or suppressed by transcription factors. Ibn other words, the association is not causal, but a correlation. See here for further explanation.
  • The difference between identical twins is due to different epigenetic markers caused by their different life experiences. Again, this is not known; any associations may be correlations rather than the markers causing differential expression of identical genes in identical twins.
  • Inherited epigenetic markers (methylated DNA bases in genes) can be passed on to future generations, so that an environmentally-induced change in the genetic material, affecting morphology, physiology, or behavior, could be passed on to future human generations.  While this is the case for up to two or three generations in some species, no long-term changes of that type are known (ergo “epigenetic changes” cannot be the basis of a kind of Lamarckian evolution). Further, this is not known to happen in humans, although there is some evidence that parental experiences can affect the offspring’s phenotype or behavior—for only a generation or two. The video contends that “Your mother’s or your father’s experiences as a child, or choices as adults, could actually shape your own epigenome.” There is not a SHRED OF EVIDENCE for that claim. (Note that the video’s title is “How the choices you make can affect your genes.”)
  • A healthy diet, exercise, and avoidance of exposure to contaminants can give you a healthy epigenome. Shades of Chopra! Can “epigenetic cleanses” be far behind? That’s just complete hogwash.
  • Epigenetics can explain the origin of cancer, heart disease, mental illnesses, and other diseases. In other words, the study of epigenetics can help us cure disease.  In fact, there’s no evidence for this contention either. In a piece in the July 1 New York Times, “Growing pains for field of epigenetics as some call for overhaul“, Carl Zimmer highlights a new paper in PLoS Genetics by Ewan Birney, George Davy Smith and John M. Greally (reference and free link below)—a paper that severely criticizes the evidence that diseases can be caused by epigenetic modifications. It entirely possible that such modifications, if they’re real, are likely to be the consequences of disease.

Zimmer’s piece says this:

In May, Duke University researchers claimed that epigenetics could explain why people who grow up poor are at greater risk of depression as adults. Even more provocative studies suggest that when epigenetic marks change, people can pass themto their children, reprogramming their genes.

But criticism of these studies has been growing. Some researchers argue that the experiments have been weakly designed: Very often, they say, it’s impossible for scientists to confirm that epigenetics is responsible for the effects they see.

Three prominent researchers recently outlined their skepticism in detail in the journal PLoS Genetics. The field, they say, needs an overhaul.

“We need to get drunk, go home, have a bit of a cry, and then do something about it tomorrow,” said John M. Greally, one of the authors and an epigenetics expert at the Albert Einstein College of Medicine in New York.

Among other criticisms, he and his co-authors — Ewan Birney of the European Bioinformatics Institute and George Davey Smith of the MRC Integrative Epidemiology Unit at the University of Bristol in England — argue that in some cases, changes to epigenetic marks don’t cause disease, but are merely consequences of disease.

And here’s one way that this claim has been investigated:

Some studies, for example, have found that people with a high body mass index have unusual epigenetic marks on a gene called HIF3A. Some researchers have suggested that those marks change how HIF3A functions, perhaps reprogramming fat cells to store more fat.

If that were true, then drugs that reverse these changes might be able to help obese people lose weight. But Dr. Smith and his colleagues have found that overweight subjects experienced epigenetic changes to HIF3A only after they put on weight.

And one more caveat:

Dr. Greally and his colleagues note another source of confusion: Normal genetic variation leads some people to produce different epigenetic marks than others.

If researchers were to find that alcoholics carry an unusual epigenetic mark, for instance, that wouldn’t necessarily mean that it resulted from heavy drinking. These people may have a genetic variation that puts them at risk of alcoholism and, perhaps coincidentally, creates an unusual epigenetic mark on their DNA.

Dr. Greally said these possibilities have been neglected because scientists have been so captivated by the idea that epigenetic marks can reprogram cells.

“Since you don’t talk about anything else, you interpret the results solely through that little sliver of possibility,” he said.

He and his colleagues go so far as to claim that no published results on the links between epigenetic marks and disease “can be said to be fully interpretable.”

From what I know (and I’ll admit this isn’t my field), these criticisms are on the mark. But read the PLoS Genetics article below if you want to go deeper into this field. In the meantime, it’s just irresponsible for TED to promulgate the video below, which makes about as many mistakes as can be made in a five-minute presentation. I’d suggest that TED consider removing this misleading presentation. In the meantime, my advice to the layperson reading popular articles about epigenetics is this: unless the article is by Carl Zimmer, take everything it says with a grain of salt.

The misguided TED video:



  1. Barry Lyons
    Posted July 3, 2016 at 11:20 am | Permalink

    “Epigenetics cleanse”! Now that’s hilarious.

  2. Posted July 3, 2016 at 11:29 am | Permalink

    I predict Chopra will write a book about quantum epigenetics. Woo Squared.

  3. rickflick
    Posted July 3, 2016 at 12:13 pm | Permalink

    “epigenetics could explain why people who grow up poor are at greater risk of depression as adults.”

    Ha! I would think being poor would be enough all by it self.

  4. Mark Sturtevant
    Posted July 3, 2016 at 12:29 pm | Permalink

    Ditto on everything you say. Especially the advice to rely on Zimmer in the general press.

  5. Michael Scullin
    Posted July 3, 2016 at 12:37 pm | Permalink

    I won’t be watching the TED production because they tend to make me either cringe or angry. But I do appreciate your meticulous dissection of the program because it covers much of the current contention about epigenetics with great clarity. I saved Zimmer’s article for my file on epigenetics which includes your recent series of columns. The term is loosely kicked around by some social scientists who seem especially apt to pick up some new term and abuse it until a new term is discovered.

  6. Ben
    Posted July 3, 2016 at 1:20 pm | Permalink

    Thanks for the explanations and coverage on epigenetics.

    I tend to agree with avoiding TED talks, but this seems to be the kind of thing that is just up their line. Too many of them sound to me like multi-level marketing or Trump university presentations.

  7. Eduardo
    Posted July 3, 2016 at 1:24 pm | Permalink

    Good post. Thank you.

  8. Joseph McClain
    Posted July 3, 2016 at 1:46 pm | Permalink

    Very helpful! Thanks, Jerry.

  9. Posted July 3, 2016 at 2:05 pm | Permalink

    I should weigh in on this topic since Jerry has chosen to write not only about my dissertation topic but about a paper my upcoming postdoc boss (Smith) co-authored.

    I need to think.

  10. JonLynnHarvey
    Posted July 3, 2016 at 2:28 pm | Permalink

    On this glorious 4th of July weekend, upon hearing the phrase “Shades of Chopra”, I am going to think of these lovely photographs of Bollywood actress Priyanka Chopra wearing a variety of dark glasses.


    Sample here

  11. Posted July 3, 2016 at 3:06 pm | Permalink

    The methylation technology I know is Illumina’s 450K array, which “genotypes” sodium bisulfite-treated DNA to quantify DNA methylation for close to half a million CpG loci, capturing roughly 2% of the CpG sites throughout the human genome(1). Bisulfite conversion allows us to use genotyping technology to measure methylation. Treatment with bisulfite converts unmethylated cytosines (C) into uracils, which are subsequently converted to thymines (T), one of the four DNA bases. This allows us to treat a CpG site as if it were a C/T polymorphism, which can be genotyped and interpreted as a proxy for methylation status. However, unlike the categorical feature of genotypic calls obtained from a genome-wide association study (GWAS), what is measured at a CpG site is the percentage of cells that are methylated for the particular tissue sampled. (In my case the tissue is peripheral blood and measured cells are lymphocyte subtypes.) After bisulfite-converted DNA is amplified and fragmented, it is hybridized to two probes, referred to as Type I and Type II probes(2). The Type I and II probes quantify methylation differently, though they both obtain the standard index for methylation: beta (β), where [β = M/(M + U)] and M and U refer to the methylated and unmethylated signal intensities at a given CpG site. A series of (bioinformatic) pre-processing steps are necessary before the beta values can be analyzed.

    1) Blood is composed of various cell subtypes. The relative abundance of the different cell lineages can confound the measurement of methylation, such that observed differences in methylation are likely explained not by the exposure of interest but by the differences in subtype abundance. As such, cell composition has to be addressed.
    2) Probe filtering has to be done to remove poor-performing probes, ones that fail to bind to the bead chips. Two approaches are typically used in the literature to filter poor-performing probes: knowledge of bead counts and use of detection P-values(3).
    3) The underlying genetic sequence can influence the methylation status of CpG loci; single nucleotide polymorphisms (SNPs) can impact hybridization, and genetic variation, similar to population substructure in genetic studies, can account for differences in methylation.
    4) A limitation of genotyping bisulfite-converted DNA is that it greatly reduces the genome down from four letters (A, T, C, G) to, essentially, three (A, T, G) with only 3.5% of C’s remaining detectable. As a result, the specificity of probes designed to hybridize to unconverted DNA is reduced. Up to 25% of 450K probes cross-react (cross-hybridize) to sites other than those for which they were designed, which is particularly problematic, as the measurement of methylation at these sites then represents a composite of methylation intensities from CpG sites across the genome(3). Cross-reacting probes should be removed to avoid spurious association.
    5) When a probe hybridizes, it incorporates a labeled dideoxynucleotide triphosphate (ddNTP), which fluoresces. The ratio of fluorescence from the methylated and unmethylated beads is used to determine the signal intensity for the beta value. However, a certain amount of background fluorescence exists between arrays and biases the beta values if left uncorrected. Background correction corrects this bias by estimating the true signal from the foreground, subtracting out the background mean intensity(4).
    6) After background correction, the overall distribution of the beta values still differs between the two probe types, owing to the different chemistries used by them. The betas need to be rescaled-—normalized.
    7) Methods for correcting batch effects are also needed.

    Owing to the complexity here, it is incredibly challenging to compare the results of various methylation studies. However, I don’t think that should stop us from doing methylation research.

    1. Michels, K. B. et al. Recommendations for the design and analysis of epigenome-wide association studies. Nat. Methods 10, 949–955 (2013).
    2. Pidsley, R. et al. A data-driven approach to preprocessing Illumina 450K methylation array data. BMC Genomics 14, 1–10 (2013).
    3. Dedeurwaerder, S. et al. A comprehensive overview of Infinium HumanMethylation450 data processing. Brief. Bioinform. 15, 929–941 (2014).
    4. Triche, T. J., Weisenberger, D. J., Van Den Berg, D., Laird, P. W. & Siegmund, K. D. Low-level processing of Illumina Infinium DNA Methylation BeadArrays. Nucleic Acids Res. 41, 1–11 (2013).

  12. Mark R.
    Posted July 3, 2016 at 4:06 pm | Permalink

    A lot of the points TAD got wrong were the same points Neil deGrasse Tyson got wrong on his recent video on epigenetics. As pointed out by PCC(E) here

    • Mark R.
      Posted July 3, 2016 at 4:09 pm | Permalink

      TAD? TED

  13. gravelinspector-Aidan
    Posted July 3, 2016 at 6:52 pm | Permalink

    Shades of Chopra! Can “epigenetic cleanses” be far behind? That’s just complete hogwash.

    After build up like that, I look forward to the release of the forthcoming line of “Ceiling Cat Epigenetic Cleansing Colonic Irrigation Washes”.
    NOW I understand the jaunty script on the new boots – it’s the company logo. Instead of everyone using “Jerry & Ben” ice cream, we’ll get “Coyne & Jerry’s Epigenetic Colonic Washes”!
    Next step – World Domination (fast).

  14. peepuk
    Posted July 4, 2016 at 4:39 am | Permalink

    When we see some moral claims in these stories we can be sure that this part of the story is wrong.

    TED is proof that translating science into an entertaining stories is difficult. Science is very dull except maybe for the experts in their fields; some pimping up for the general public is always needed.

  15. Posted July 4, 2016 at 9:24 am | Permalink

    A one-generation effect on the germline epigenome can be an awfully powerful source of developmental pathology. I’m not sure why you or Greally et al scoff at that biological reality. I was doused in synthetic steroid hormone drugs as a fetus, all while my oocytes were being reprogrammed. My children suffer from markedly and idiopathically abnormal neurodevelopment (there is zero mental disorder in my or husband’s families). The urgent question facing science must be “did epigenomic dysregulation of the oocyte precipitate serious pathology?” and not this facile, cackling piling-on I see from you and Greally, Birney. Btw yes I have encountered numerous families with my same exact exposure and pathology story.

    • Posted July 4, 2016 at 9:31 am | Permalink

      Nobody is arguing with the fact that giving pregnant mothers chemicals or drugsw can change the offspring. What we are saying, and you offer NO evidence to the contrary, is that there’s simply no evidence for any epigenetic reprogramming of the genome that is responsible for this stuff–OR for the heritability of parental alterations (not drugs taken when pregnant) to the next generation.

      You haven’t followed this controversy here–that is obvious–and so it is not “facile cackling piling-on”. Did you read the many famous people working on epigenetics who also attacked claims like those in the TED video.

      I’m sorry, but you haven’t any evidence that your own steroid hormone dose caused “reprogramming of the germline epigenome.” Sorry, but you’re the facile and cackline one, making claims for which there is no evidence.

      And, of course, you’re rude.

  16. Posted July 4, 2016 at 12:20 pm | Permalink

    I’ve always thought that so-called identical twins were different because of the fact that any organism is sufficiently large to have random portions to its development. For example, in mammals slightly different concentrations of relevant development proteins (because concentrations are never uniform), etc.

    Is this wrongheaded?

  17. somer
    Posted July 5, 2016 at 4:29 am | Permalink

    In such a complex and highly topical field tripe readily gets traction in the media – who wouldn’t want to have an epigenetic life – either to pass blame or to hope for miracles?

  18. Posted July 5, 2016 at 6:38 am | Permalink

    Reblogged this on My Selfish Gene and commented:
    I wish there was more thoughtful commentary like this covered by news sources. But in the age of TED, SEO, and the Huffington Post I am not holding out hope. Enjoy a short read to straighten out most of what you’ve heard about epigenetics.

  19. Posted July 5, 2016 at 4:23 pm | Permalink

    I think it would be interesting if biologists make a brief video explaining what’s epigenetics really is (and what it’s not, such as the common mistake as thinking it has something to do with Lamarckian evolution).

    It’s very easy to that this kind of video; almost all smartphone today have a great camera. It would only it need someone to edit the video, putting all pieces together, maybe with subtitles, saying who is who/field of work, and so on.

  20. Posted July 6, 2016 at 12:05 am | Permalink

    While the statement “It is entirely possible that such modifications, if they’re real, are likely to be the consequences of disease” is true, the knowledge that is gained from well-designed studies, ones that tease out the timing of biomarkers of intermediate effect, is critical not only to causality but, potentially, to the early identification of at-risk populations and which patients, once they have cancer, are likely to respond to treatment.

    I’ll provide you with one paper that suggests that methylation signatures function as phenotypic markers of susceptibility to cancer (1). And this may be true without getting into whether methylation is up or downstream from the activity of transcription-factor binding. Still so, a phenotypic marker doesn’t imply causality, though it leaves the possibility open that changes in methylation are risk factors for cancer. An intriguing possibility is that factors associated with both methylation and genomic instability, potentally chromosomal aberrations or telomere length, or something else, are what’s causal.

    This paper I’m citing is now ancient (2008) but a bedrock in cancer epidemiology. Among its coauthors is the now head (Stephen Chanock) of the Division of Cancer Epidemiolgy and Genetics, the basic science arm of cancer epidemiology, at the National Cancer Institute

    1. Moore, L. E. et al. Genomic DNA hypomethylation as a biomarker for bladder cancer susceptibility in the Spanish Bladder Cancer Study: a case-control study. Lancet Oncol 9, 359–366 (2008).

  21. Posted July 6, 2016 at 10:14 pm | Permalink

    Okay, I’m dominating this thread, but I keep thinking of more stuff to add! I thought about adding this comment to other epigenetics posts, but the comments for them are already closed.

    I think I know how to explain what people are trying to communicate regarding the place of methylation relative to transcription factors. Using KRAS, an oncogene activated in many colorectal cancers, as an example, Serra et al. (2014) found that KRAS stimulates enzymes that inhibit the degradation of the transcription factor (ZNF304), which then increases in concentration and recruits a co-repressor complex, including several proteins and a DNMT1 (a DNA methylase). The complex was subsequently targeted to the tumor suppressor gene, INK4-ARF, methylating and repressing it. According to Struhl (2014), loss of methylation occurs if any of the associated proteins are inhibited(1,2). Whilst this supports the primacy of transcription factors in directing methylation, there is other evidence, albeit older, that suggests that a hit to the transcription factors isn’t the only possible pathway to change methylation in such a way as to promote carcinogenesis. Eden et al. (2003) introduced a hypomorphic allele at DNMT1 (causing hypomethylation) in NF1 +/- and p53 +/- (cancer-prone) mice and compared the ages they developed sarcomas to the ages of tumor initiation in their methylation-capable littermates. The mice with the underexpressing DNMT1 developed sarcomas sooner. They then examined the rate of loss of heterozygosity (LOH) among primary embryonic fibroblasts by developing an assay to score the fraction of LOH (NF1 -/- and p53 -/-) among heterozygotes and found that the rate of LOH increased in hypomethylated cells. They then calculated the frequency of LOH at chromosome 11 (location of NF1 and p53 in mice; it’s chromosome 17 in humans) in hypo and methylated cells and found that the frequency was higher in the hypomethylated cells(3). These lines of evidence point to the downregulation of DNMT1 as being a plausible carcinogenic mechanism.

    1. Serra, R. W., Fang, M., Park, S. M., Hutchinson, L. & Green, M. R. A KRAS-directed transcriptional silencing pathway that mediates the CpG island methylator phenotype. Elife 1–22 (2014).

    2. Struhl, K. Is DNA methylation of tumour suppressor genes epigenetic? Elife 1–3 (2014).

    3. Eden, A., Gaudet, F., Waghmare, A. & Jaenisch, R. Chromosomal instability and tumors promoted by DNA hypomethylation. Science 300, 455 (2003).

%d bloggers like this: