Alice is wrong: at every size she is still Alice, and nearly every change came from something she ate or drank. Carroll never says what the cake or the bottle changed inside her.

Figure 1. Left: Alice holds up the “DRINK ME” bottle. Right: after the “EAT ME” cake, “opening out like the largest telescope that ever was.” Illustrations by John Tenniel (1865), public domain, via Wikimedia Commons (left, right).

A bottle in the bathroom cabinet works the other way: its childproof cap changes none of the pills inside, only who gets to them.

Feed a bee larva royal jelly and it grows into a queen instead of a worker, on the same DNA it started with. If its genes are a cookbook, the jelly writes no new recipes. It changes the notes in the margins that say which recipes get read: a layer of marks on DNA called the epigenome.

It is your next checkup, at a new clinic, and your pen stalls over the intake form’s family history, at the box for diabetes. If your parents moved countries, or you did, your DNA made the trip unchanged. Your odds of getting Type 2 Diabetes may not have. If your family is Asian American, the American Diabetes Association says to consider a blood sugar test at a body mass index of 23, not 25. Nearly one in four American adults with diabetes does not know they have it. If your mother is one of them, you leave the box blank, and your doctor loses one reason to test you.

Why would a move change the odds and leave the DNA alone? That gap is what a paper I worked on is about. Close to midnight, seven years ago, a dozen drafts into the manuscript, I have been stuck on one paragraph for twenty minutes.

The paragraph makes the jelly’s claim about people: two groups who share an ancestry, and with it most of their DNA. One group moved across an ocean and one stayed home (Figure 2). People with diabetes carry different chemical tags on a gene called TXNIP than people without it. The ones who moved carry a few differences of their own. I cannot vouch for the paragraph’s statistics, because I cannot picture what a tag does to a gene whose letters never changed.

Figure 2. The paper’s two cohorts: Filipinos born and living in the Philippines, who stayed, and the overlap, born there and living in the United States, who moved. Same ancestry, a different country. Drawn for this site’s Research page.

On this paper, mine is the tenth of sixteen names: the statistician. I checked the analysis with Dominic, the lead author, sorting real differences from chance. Tonight my whole job is one comment in the margin, on whether the statistics hold.

I know the basics of the double helix and how a gene is copied into RNA. What I do not have is the vocabulary everyone else on this list uses every day, starting with what a promoter region does.

The little I have picked up from this list says chemical tags sit near the front of a gene and turn it down. The paragraph on my screen puts the tags inside the gene, near the gene’s far end, and I cannot see what a tag would be doing there. I have read the paragraph so many times tonight, the words have started to come apart. My eyes go a little heavier each time I reach “differentially methylated.”

I do not remember deciding to close them.

Someone is already waiting when I open them again, and it feels like Wonderland. The Hatter is there, in a hat two sizes too large, holding a full teacup. His voice is David Attenborough’s, hushed and marveling: “Stand back and you see an entire field. Stand close and you carry a single bead.”

Who Puts the Caps On

Six billion tiny sprites hold hands in two long twisted lines across a field. Each presses a bead to its chest, stamped T, C, G or A. A second field lies beside it: same sprites, same order, letter for letter.

The Hatter: Every cell in a human body, with only a few exceptions, carries an identical copy of the genome. A neuron and a liver cell do different work on identical instructions, so something else decides which parts get followed. A verse, here, means one gene, the stretch a courier stops to copy.

Back among the sprites, I shrink down until the double line towers on both sides of me. Something small goes from sprite to sprite with a satchel: an imp, far quieter than the courier imps that copy verses. This one goes by Hush, the keeper of the caps.

Hush carries no beads and threads nothing. Its satchel holds tiny knitted nightcaps, soft and pointed, with a tassel on the end. Hush caps only the C sprites, and only a C with a G next in line, choosing which ones keep working today and which go to sleep.

A nightcap does not unlink a sprite’s hands: T still finds A, and C still finds G, with the cap riding on top. Hush tugs it down over a sprite’s eyes, and the sprite yawns once and drifts off, mid-chain, still holding on.

The caps that put a verse to sleep go on its flag, the short run of sprites just before the verse that tells a courier where to start reading. A courier flying past finds the flag’s sprites asleep, still holding their beads but giving no sign to stop. The courier flies on, and the verse goes unread, a bottle whose cap no courier can turn.

The Hatter: The flag has a name: a gene’s promoter. A courier builds its thread in one direction only, from the end biologists call 5’ toward the end called 3’. So the flag lies upstream, on the 5’ side of its verse. Caps on a promoter keep the reading machinery from starting there, and that is how they silence the gene.

Hush caps the same flags every visit, cell after cell: a handful of verses asleep, the rest wide awake. Most of its caps go elsewhere, between verses and down the middle of awake ones, where the caps stop a courier from starting partway down. A courier already reading runs straight over them.

The Hatter: This is called methylation. Hush stands for an enzyme, a DNA methyltransferase, the protein that lays the caps. The nightcap is a methyl group, a small chemical tag attached directly to DNA, in people almost always on a C followed by a G. So a protein decides which DNA gets read, and control runs back up the chain that turns DNA into protein.

But who tells Hush which flags to cap?

What Moves the Caps

In a growing embryo, Hush’s caps help decide which verses a liver cell will read and which a neuron will. Later, other signals move the caps: a nutrient the body went short of in the womb (Figure 3), a hormone running high for years, a long stretch of chronic inflammation. Each one changes where Hush lays its caps, as royal jelly does in a bee larva.

epigenetics, n.

▶

“The study of how DNA interacts with the multitude of smaller molecules found within cells, which can activate and deactivate genes.”

Some caps are set before there is an embryo, in the egg or the sperm, by which parent a copy came from. On one stretch of chromosome 15, some genes are read only from the father’s copy and one only from the mother’s. The other copy stays silent, a second bottle capped before birth. Lose that stretch from the father’s copy and a child has Prader-Willi syndrome. Lose it from the mother’s, and the child has Angelman syndrome, a different disorder of development from the same missing letters.

The Hatter: Some marks come from outside the body. In a smoker’s blood, a site on a gene called AHRR loses its caps. After the last cigarette, most smoking-linked sites drift back within about five years, and a few still lack their caps decades later. For someone who has quit, the number to count is the years since the last cigarette.

Figure 3. Same DNA, different caps. Two genetically identical agouti mice. In the brown one, methylation keeps the agouti gene quiet; in the yellow one, it does not, and the mouse grows fat and prone to diabetes. When their mothers eat extra folic acid and other methyl donors in pregnancy, more of the pups come out brown. Mouse photo by Randy Jirtle and Dana Dolinoy (2007), CC BY 3.0, via Wikimedia Commons.

Caps can also land where they do harm. Some colon tumors show the pile-up of slips a broken inspector leaves, with no inherited cause. Most of those tumors carry caps on the flag of MLH1, one of the genes that build the inspector. Every letter of MLH1 is in place: the cap is on the inspector’s own bottle. In the lab, taking those caps off put the inspector back to work.

I pull back until both double lines fit in view, the same six billion sprites in each. In one, a verse lies wide open. In the other, the same verse sleeps under Hush’s caps with every letter in place. The sprites’ amber eyes blink out before I can see what a move across an ocean would change.

Where the Caps Come Off

At one in the morning, I wake at my own desk, the cursor blinking in the empty comment box. My cheek carries the print of the keyboard’s edge. The paragraph that stopped me is still there, unchanged. But now I can picture it.

Two cohorts stand behind this paper, an ocean apart: 198 adults born in the Philippines and still living there, and 167 Filipinos who moved to the United States. The two groups share an ancestry. Their lives since have split. Blood from both is read on a chip that checks about 850,000 sites at once.

Researchers call this design a natural experiment: a real-world split that works like a controlled study, though nobody assigned anyone to a group. Migration did the assigning. With ancestry held steady, whatever differs afterward in methylation becomes a candidate for something the years apart did. The design cannot say which thing in those years did it: food, work, income, a health system, a winter (Figure 4).

▶

Figure 4. The Dutch Hunger Winter of 1944–45, the best-known natural experiment in epigenetics, from MinuteEarth (2015), YouTube embed. The frightened mice belong to the video’s later, still-debated claim: marks passed from parents to offspring.

Type 2 Diabetes is a good disease to test this way. Genes explain only part of who develops it, and among Filipinos it is more common in those who moved to the United States.

After quality checks, about 716,000 of the chip’s sites were tested. By my own arithmetic, at the usual one-in-twenty bar, about 36,000 of them would look like finds by chance alone. My part was the correction that keeps chance hits to a small share of what gets called real. After the correction, 177 sites were left standing as tied to diabetes, each shifted, on average, by about one cap in twenty.

Three of the 177 fall on TXNIP, the gene from the paragraph I was stuck on. Rising blood sugar pushes TXNIP to make more of its protein, and too much of it damages the cells that make insulin (Figure 5). Whether the three shifts come before the diabetes or after it, a single blood draw cannot say.

Figure 5. TXNIP, the protein rising blood sugar calls up, as AlphaFold predicts it folds: two domains in blue and yellow and a loose grey tail. In people with diabetes the caps on its gene had shifted, but every letter, and so this shape, stayed the same. Image by Leskomitch42 (2022), from AlphaFold (UniProt Q9H3M7), CC BY-SA 4.0, via Wikimedia Commons.

The three sites lie inside the TXNIP verse, near its far end, right where the paragraph had put them. In the people with diabetes, those three sites had lost their caps: the ones Hush lays along the middle of an awake verse. Where the caps came off, the paper saw more copying of the verse’s tail, as if a courier had found a flag where none belonged. The paper’s title calls this spurious transcription initiation. That false start is the part my coauthors understood long before I did, the part I needed six billion sprites and an imp named Hush to finally hold onto.

On the intake form’s family history, I tick yes beside diabetes. My mother’s whole family has diabetes, except her. My grandfather on her side died of it. So did one of her sisters, from complications. She was on dialysis. Each time I got pregnant, the first time and the second, my doctors checked that history. It puts me at risk for gestational diabetes, the kind that starts during pregnancy. I had a stake in this before the paper, and after it. I also made the same move as the paper’s second cohort, from the Philippines to the United States. The paper says the move can count too.

The two fields were hiding a second thing. One stood for the cohort that stayed, one for the cohort that left. The two cohorts’ caps had landed in different places. Where diabetes and the move overlapped, the analysis found three more sites, near genes called MAP2K7, PRMT1 and ADM5.

I did not know that night whether the comment I left would be any good. I know now. The paper published at the end of 2019. TXNIP made it past every round of review, and the sentence that stopped me that night has been in print ever since. The dream went fuzzy by morning, and I did not think about it again for a long time.

Seven years later, it is the field that came back, not the paragraph, not the comment, not even the paper.

A Closing Invitation. Like Alice at every size, the same words can land differently, and the same DNA can be read differently. Right now, say this out loud seven times, landing on a different word each time:

I never said she stole my money. (Someone else did.)
I never said she stole my money. (Not once, not ever.)
I never said she stole my money. (I only implied it.)
I never said she stole my money. (Her brother did.)
I never said she stole my money. (She borrowed it.)
I never said she stole my money. (She took yours.)
I never said she stole my money. (She stole my heart.)

Which version would you least want your best friend to hear?

Hush never deletes a gene, only caps it. The notes in the cookbook’s margins were caps all along.

Worth remembering Save as image

Off is not gone. A chemical cap can silence a gene without changing a letter, and smoking or a famine can move the caps. So quitting counts: within five years, most caps smoking moved are back.

Where This Practice Came From

Waddington’s 1942 paper, three pages in Endeavour, is where the word epigenetics starts. The field and the sprites first appeared in the walkthrough of the double helix, and the courier imps in the walkthrough of copying a gene.

Intellectual Honesty Note. The sprites, the imps, the field, the nightcaps and the rest of the staging are invented, built to make old, well-tested science easier to hold onto, never to replace it. So are the dream, the Hatter and his David Attenborough voice, and the night at the desk is dramatized. Each mechanism the sprites act out is textbook biology, and the Hatter’s asides name it plainly so the metaphor never carries it alone. The video in Figure 4 goes further than this piece, into marks said to pass from parents to children and grandchildren, which is still debated; this piece leans only on its account of the Dutch famine as a natural experiment. The false start at TXNIP is the paper’s reading of the marks around those sites, not a courier caught in the act.

References

Albao, D.S., Cutiongco-de la Paz, E.M., Mercado, M.E., Lirio, A., Mariano, M., Kim, S., Yangco, A., Melegrito, J., Wad-asen, K., Gauran, I.I., Francisco, M.A., Santos-Acuin, C., David-Padilla, C., Murphy, E.J., Paz-Pacheco, E. and Seielstad, M., 2019. Methylation changes in the peripheral blood of Filipinos with Type 2 diabetes suggest spurious transcription initiation at TXNIP. Human Molecular Genetics, 28(24), pp.4208-4218.

Alberts, B., Heald, R., Johnson, A., Morgan, D., Raff, M., Roberts, K., & Walter, P. (2022). Molecular Biology of the Cell (7th ed.). W. W. Norton.

American Diabetes Association Professional Practice Committee. (2022). 2. Classification and diagnosis of diabetes: Standards of Medical Care in Diabetes, 2022. Diabetes Care, 45(Suppl. 1), S17–S38.

Carroll, L. (1865). Alice’s Adventures in Wonderland. Macmillan.

Centers for Disease Control and Prevention. (2023). National Diabetes Statistics Report. https://www.cdc.gov/diabetes/data/statistics-report/index.html

Guerrero-Bosagna, C. (2016, June 27). What is epigenetics? [Video]. TED-Ed. YouTube. https://www.youtube.com/watch?v=_aAhcNjmvhc

Heijmans, B. T., Tobi, E. W., Stein, A. D., Putter, H., Blauw, G. J., Susser, E. S., et al. (2008). Persistent epigenetic differences associated with prenatal exposure to famine in humans. Proceedings of the National Academy of Sciences, 105(44), 17046–17049.

Herman, J. G., Umar, A., Polyak, K., Graff, J. R., Ahuja, N., Issa, J. P., et al. (1998). Incidence and functional consequences of hMLH1 promoter hypermethylation in colorectal carcinoma. Proceedings of the National Academy of Sciences, 95(12), 6870–6875.

Holliday, R. (2006). Epigenetics: a Historical Overview. Epigenetics, 1(2), 76-80.

Jirtle, R., & Dolinoy, D. (2007). Agouti mice [Photograph]. Wikimedia Commons. https://commons.wikimedia.org/wiki/File:Agouti_Mice.jpg

Joehanes, R., Just, A. C., Marioni, R. E., Pilling, L. C., Reynolds, L. M., Mandaviya, P. R., et al. (2016). Epigenetic signatures of cigarette smoking. Circulation: Cardiovascular Genetics, 9(5), 436–447.

Knoll, J. H., Nicholls, R. D., Magenis, R. E., Graham, J. M., Lalande, M., & Latt, S. A. (1989). Angelman and Prader-Willi syndromes share a common chromosome 15 deletion but differ in parental origin of the deletion. American Journal of Medical Genetics, 32(2), 285–290.

Kucharski, R., Maleszka, J., Foret, S., & Maleszka, R. (2008). Nutritional control of reproductive status in honeybees via DNA methylation. Science, 319(5871), 1827–1830.

Leskomitch42. (2022). Alpha arrestin structure [Image; TXNIP predicted by AlphaFold, UniProt Q9H3M7]. Wikimedia Commons. https://commons.wikimedia.org/wiki/File:Alpha_Arrestin_Structure.jpg

MinuteEarth. (2015, November 18). Epigenetics: Why inheritance is weirder than we thought [Video]. YouTube. https://www.youtube.com/watch?v=AvB0q3mg4sQ

Neri, F., Rapelli, S., Krepelova, A., Incarnato, D., Parlato, C., Basile, G., et al. (2017). Intragenic DNA methylation prevents spurious transcription initiation. Nature, 543(7643), 72–77.

Tenniel, J. (1865). [Illustrations]. In L. Carroll, Alice’s Adventures in Wonderland. Macmillan.

Waddington, C. H. (1942). The Epigenotype. Endeavour, 1, 18–20.

Waterland, R. A., & Jirtle, R. L. (2003). Transposable elements: targets for early nutritional effects on epigenetic gene regulation. Molecular and Cellular Biology, 23(15), 5293–5300.

Near the edge of the frame, a few cells hang straight down, long and bumpy, like peanut shells made of wax (Figure 1). In each one, a larva floats on a pool of thick white jelly, far more than it can eat.

Figure 1. Two queen cells on the comb, one turn of a loop. A larva fed royal jelly in each will come out a queen, and the queen will lay the eggs of the workers who feed the next one. Photograph by Piscisgate (2014), Wikimedia Commons, CC BY-SA 4.0.

Any female larva in the hive can grow into a queen or a worker. Her genes do not decide which. The larvae in the hanging cells become queens. A queen lives for years and can lay more than a thousand eggs a day. A worker lives a few weeks in summer and usually lays none.

Food decides which. The white jelly is royal jelly (Figure 2), a rich food made in glands in the workers’ heads. The workers feed it to a few larvae for as long as they are larvae. The rest get a plainer mix with pollen and honey in it, and become workers. Food cannot rewrite a gene. So what does it change?

▶

Figure 2. How royal jelly makes a queen, from SciShow (2015), YouTube embed. The video credits one jelly protein, royalactin. A later team could not reproduce that result.

Think of a larva’s genes as a cookbook. Each gene is a recipe for one molecule the body uses, and every larva in the hive carries the same book. Cells can stick small chemical marks onto some of the recipes, like notes in the margin that say “skip this one.” Biologists call them methylation marks. They help decide which recipes the cell cooks and which it skips.

In 2008, a team at the Australian National University, in Canberra, tested whether those notes were what the jelly changed. The team switched off one gene in newly hatched larvae: the recipe for the enzyme that writes the notes. Then they fed the larvae a worker’s diet. Most of them came out as queens, with full ovaries. The team called the result “a royal jelly-like effect.” Each larva kept the cookbook it was born with. What changed was which recipes got cooked.

Which bee each larva became depended on what its cookbook was connected to: the notes, the food, and the workers who chose which larvae to feed. Any cookbook raises five questions: what it counts, what it keeps, what it answers, what it misses, and what it cannot decide.

Say you mail a spit sample to a DNA testing company, and the report comes back: you probably cannot digest milk as an adult. You start drinking your coffee black, bitter at the first sip. You stop buying cheese, pay extra for lactose-free milk and read every label in the shop. Your partner switches to oat milk so the fridge needs only one carton. By your next checkup, you have gone months without an easy source of calcium.

Here is how such a report could be wrong. For most people in the world, the body puts the recipe for digesting milk away after childhood. Some people carry a one-letter difference in the cookbook that keeps the recipe in use. In Europe, most people who have one share the same spelling. Many families who kept cattle in East Africa carry a different spelling that keeps the same recipe in use. Say the test looks only for the European spelling. A person who carries the East African one gets a report that finds nothing it knows and says the recipe is closed. That person gave up milk for nothing. The test read the right page, looked for the wrong spelling, and never asked where the family came from. Systems thinking has a question for any test result that costs someone milk, cheese and calcium: what else is that result connected to?

What the Cookbook Counts

Counting human genes started as a bet. In May 2000, at Cold Spring Harbor, the geneticist Ewan Birney opened a pool on how many genes human DNA would turn out to hold. Biology was about to get its first full cookbook for a person: every gene, every recipe (Figure 3).

Figure 3. The human genome printed as books at the Wellcome Collection, in London: a volume or more for each chromosome, from 1 to 22, then X and Y, in type almost too small to read. Photograph by Russ London (2010), Wikimedia Commons, CC BY-SA 3.0.

The guesses ran from about 20,000 to more than 200,000. In 2003, the count came in at 24,847. The closest guess, 25,947, belonged to Lee Rowen, who worked at the Institute for Systems Biology. She won $600 and a signed copy of The Double Helix.

Later counts settled lower, near 20,000. The worm C. elegans, about a millimeter long, has a little over 20,000 genes of its own. A person and a worm carry cookbooks of about the same size, so the number of recipes could not be what makes a person a person. In Donella Meadows’ definition of a system, the parts matter less than the rules that connect them. Here the parts are the genes, and the rules decide which recipes get cooked, when, and how much. Two kitchens with the same book can serve very different dinners.

A full gene count is one example of omics: measuring one layer of a living thing all at once. Each layer takes the suffix -ome. Two of them explain the bees.

In this piece’s kitchen, the genome is the cookbook, and the epigenome is the notes in its margins. The genome is the layer the walkthrough of DNA, transcription, and translation takes apart. The epigenome is the layer the natural experiment on methylation follows.

The book also closes some recipes on its own schedule, and a single letter can keep one open for life.

What the Cookbook Keeps

Milk is the recipe most bodies put away. Babies make their own enzyme for lactose, the sugar in milk. The enzyme is called lactase, and in most mammals the body stops making it after weaning. A short stretch of DNA near the gene closes the recipe after childhood. In about a third of adults alive today, the recipe stays open. The European spelling is a single changed letter in that stretch, and it keeps lactase coming for life.

The East African spellings are other one-letter differences nearby, and each arose on its own. In 2007, the geneticist Sarah Tishkoff and her colleagues traced the most common of them, carried by many herding peoples in Kenya and Tanzania. It had spread fast, over roughly the last seven thousand years.

Why keep that page open? Herds. Herds put fresh milk within reach. Adults who carried the spelling could digest the milk, got more food from it, and more of their children lived to carry the spelling too. The more people could drink milk, the more a herd was worth, and herding spread further. In Meadows’ terms for feedback, that is a reinforcing loop: a change that feeds itself. Even this loop has a ceiling. A herd can grow only as large as the grass and water around it allow. The cookbook kept the page. The herds outside made it worth keeping.

A DNA sample can show which spelling someone carries. It cannot show the herds that made the spelling worth carrying, or the grass and water that capped the herds. So a kit report that rules out milk earns one email to the company within the week: which spellings did the test look for? A parent who skips the email may cut milk from the children’s plates too, on a test that never looked for the spelling the family carries. A parent who already cut the milk can still ask.

What the Cookbook Answers

The herds and their milk changed the book over thousands of years. The body asks the book questions every day, and it answers within hours.

In 2012, Romain Barrès and his colleagues put eight sedentary men on an exercise bike twice, at least a week apart. One ride was easy, at 40 percent of each man’s maximum effort, and one was hard, at 80 percent. Both rides stopped once the rider had burned 400 calories. With a needle, the researchers took small plugs of thigh muscle before each ride, right after it, and three hours later.

One recipe they read was PGC-1α, which tells a cell to build more of its energy supply. Right after the hard ride, the notes on PGC-1α had dropped to about half their resting level. Three hours later, the muscle was cooking that recipe several times as much as at rest. After the easy ride, the notes barely moved, yet the recipe still got cooked more. The notes are one rule among several.

That energy supply is mitochondria (Figure 4), the parts of a cell that burn fuel. Over months of training, more mitochondria mean more capacity for the next ride.

The ride changes the notes. The notes change which recipes get copied out to the counter, a layer called the transcriptome. The copies change which dishes get made, the proteome. The dishes change what the body can do on the next ride. The cookbook answers the ride, and the answer shapes the next question.

▶

Figure 4. What more mitochondria do for a cyclist, from Pedaling to the Top (2025), YouTube embed. Thumbnail: two mitochondria in lung tissue, electron micrograph by Louisa Howard, public domain.

To see the answer, the team had to sample the muscle three times in one day. Many omics studies draw one tube of blood from each person, once. Peter Senge defines systems thinking as seeing “patterns of change rather than static ‘snapshots.’” One tube is a snapshot. It can show the notes in the margins, but not which ride wrote them.

What the Cookbook Misses

Reading 5.4 million genomes still left out the meals. In 2022, a study led by the geneticist Loïc Yengo pooled those genomes. It found 12,111 small spellings tied to height, each with a tiny effect. Together, in people of European ancestry, they account for about 40 percent of how much heights differ. In other ancestries, the share falls to about 10 to 20 percent. Part of the 60 percent the spellings miss is genetic and not yet mapped.

The rest sits on the plate. In the 1800s, Dutch soldiers averaged about 165 centimeters, shorter than men in most of Europe. Dutch men are now the tallest in the world. They stand about 20 centimeters taller, most of a head above those soldiers. A cookbook cannot change that fast. Food, health, and living conditions can.

Say a youth basketball coach ranks twelve-year-olds by a height score built from those 12,111 spellings and cuts the bottom three. A parent shown the ranking can ask which people the score was built on. The score says less for players outside the ancestry it was built on, and nothing about what their childhood meals will add.

What the Cookbook Cannot Decide

A cell can hold a recipe and still never cook it.

Malaria kills close to 600,000 people a year, most of them young children in Africa. One of the main malaria drugs is artemisinin (Figure 5), which comes from a plant called sweet wormwood. The supply of wormwood swings. A short harvest sends the price up, so more farmers plant wormwood. Once their crops flood the market, the price falls, and the next shortage begins.

▶

Figure 5. A short history of malaria drugs, from NPR’s Skunk Bear (2012), YouTube embed.

In 2003, a team led by the chemical engineer Jay Keasling, at Berkeley, looked for a steadier source. They copied recipes from baker’s yeast and one from wormwood into E. coli, and the bacteria made an early form of the drug. The team then moved the work into yeast itself. In April 2013, the drug company Sanofi started making artemisinin with yeast in Garessio, in northern Italy. That year, the factory made about 8 percent of the world’s supply.

By 2015, farmers had a great year. So much wormwood grew that artemisinin from the plants sold for less than $250 a kilogram. Sanofi needed $350 to $400 to break even. That is at least $100 a kilogram above the plants. Sanofi’s rivals would not buy from it, and the world wanted no more of the drug than before.

“If that price is already very low and there’s a bumper crop,” Keasling said, “there’s no reason to fire up a fermenter.” The fermenter, the steel tank where the yeast grows, stood idle. The yeast still had the recipe. The price of wormwood decided.

That is the edge of every cookbook. No page inside the yeast held the price of wormwood. The jelly, the herds, the ride, the meals and the market all came from outside the book, and each one decided what got cooked.

A test report that says to give up a daily food has read one book. A friend holding such a report can ask three things first. What did the test look for? Which people was it built on? What else is the result connected to? The larva’s cookbook could not say which bee it would become. The jelly in its cell could.

A Closing Reflection. The cookbook counted, kept, answered and missed, and it did not decide. The jelly, the herds, the ride, the plate and the price of wormwood each changed which recipes reached the table, and none rewrote a recipe.

1. Spot the idle fermenter. Right now, look around for something you paid for that still works and stands unused: a bread maker in the cupboard, an app on the phone in your hand. It still has its recipe. What around it changed?
2. Read the margin notes. Think back to one habit that changed this year. Something around you changed first: a new gym on your street, an office move that added forty minutes to the commute, a friend who started walking at six. Which of your own recipes got cooked more, which got skipped, and who at home noticed first?
3. Ask which spelling. When the next report tells you to give up a food, from a DNA kit or your next checkup, ask what the test looked for before the shopping changes. One question may keep the cheese on the table. Who would give the food up with you, the one who cooks or the one who shops?

Open a packet of baker’s yeast and smell it. Split the dough: one bowl by a warm window, one in a cool corner. Look again in an hour. Same packet, same recipe. One bowl rose like a queen.

Worth remembering Save as image

You can read every page of the cookbook and still not know what is for dinner. Your genes hold the recipes; what gets cooked is decided outside the book.

Where This Came From

This piece follows systems thinking into a single cell. Konopka’s post on systems thinking (2026) started it.

Intellectual Honesty Note. The cookbook is this piece’s image, not a biologist’s term: genes as recipes, methylation marks as notes in the margin, gene variants as spellings, RNA copies as recipes copied out, proteins as dishes. A methylation mark does not always mean “skip”; on the genes in the exercise study, fewer marks went with more activity. The hive scene is built from standard beekeeping descriptions, not one hive.

The stakes paragraph’s report is hypothetical, and which spellings a kit tests depends on the kit. This piece reads lactase persistence as a reinforcing loop with a grazing ceiling. It adds the exercise loop’s last step, “more capacity for the next ride.” Its point about a snapshot not naming the ride is its own reading, not a finding. The stretch of DNA near the lactase gene is an enhancer, which raises a gene’s activity; “closes the recipe” simplifies how it is turned down. The exercise study measured single rides, not months of training. The 60 percent and the $100 gap are subtractions made here from the cited numbers, and the ride’s “about half” and “several times” are read off the exercise paper’s Figure 2. The protein step and the link from PGC-1α to mitochondria are standard physiology. The bee experiment switched the methylation enzyme off by hand; how much methylation steers caste in an ordinary hive is still debated (Oldroyd and Yagound, 2021).

References

Altun, Z. F., & Hall, D. H. (2009). Introduction. In WormAtlas. https://doi.org/10.3908/wormatlas.1.1

Barrès, R., Yan, J., Egan, B., Treebak, J. T., Rasmussen, M., Fritz, T., et al. (2012). Acute exercise remodels promoter methylation in human skeletal muscle. Cell Metabolism, 15(3), 405–411. https://doi.org/10.1016/j.cmet.2012.01.001

Buttstedt, A., Ihling, C. H., Pietzsch, M., & Moritz, R. F. A. (2016). Royalactin is not a royal making of a queen. Nature, 537(7621), E10–E12.

Enattah, N. S., Sahi, T., Savilahti, E., Terwilliger, J. D., Peltonen, L., & Järvelä, I. (2002). Identification of a variant associated with adult-type hypolactasia. Nature Genetics, 30(2), 233–237.

Gerbault, P., Liebert, A., Itan, Y., Powell, A., Currat, M., Burger, J., et al. (2011). Evolution of lactase persistence: An example of human niche construction. Philosophical Transactions of the Royal Society B, 366(1566), 863–877.

Howard, L. (n.d.). Mitochondria, mammalian lung – TEM [Electron micrograph]. Wikimedia Commons, public domain. https://commons.wikimedia.org/wiki/File:Mitochondria,_mammalian_lung_-_TEM.jpg

Kamakura, M. (2011). Royalactin induces queen differentiation in honeybees. Nature, 473(7348), 478–483.

Konopka, A. (2026, May 6). Systems thinking [Substack post]. https://thinkmicrobe.substack.com/p/systems-thinking

Kucharski, R., Maleszka, J., Foret, S., & Maleszka, R. (2008). Nutritional control of reproductive status in honeybees via DNA methylation. Science, 319(5871), 1827–1830.

Martin, V. J. J., Pitera, D. J., Withers, S. T., Newman, J. D., & Keasling, J. D. (2003). Engineering a mevalonate pathway in Escherichia coli for production of terpenoids. Nature Biotechnology, 21(7), 796–802.

Meadows, D. H. (2008). Thinking in Systems: A Primer. Chelsea Green.

National Human Genome Research Institute. (n.d.). Epigenome. In Talking Glossary of Genomic and Genetic Terms. https://www.genome.gov/genetics-glossary/Epigenome

National Human Genome Research Institute. (n.d.). Gene Sweepstakes (GeneSweep). https://www.genome.gov/virtual-exhibits/GeneSweep

National Human Genome Research Institute. (n.d.). Genome. In Talking Glossary of Genomic and Genetic Terms. https://www.genome.gov/genetics-glossary/Genome

NCD Risk Factor Collaboration. (2016). A century of trends in adult human height. eLife, 5, e13410.

NPR. (2012, December 17). Herbs and empires: A brief history of malaria drugs | Skunk Bear [Video]. YouTube. https://www.youtube.com/watch?v=IrNL27eWKOI

Oldroyd, B. P., & Yagound, B. (2021). The role of epigenetics, particularly DNA methylation, in the evolution of caste in insect societies. Philosophical Transactions of the Royal Society B, 376(1826), 20200115.

Paddon, C. J., Westfall, P. J., Pitera, D. J., Benjamin, K., Fisher, K., McPhee, D., et al. (2013). High-level semi-synthetic production of the potent antimalarial artemisinin. Nature, 496(7446), 528–532.

PATH. (2013, April 11). Sanofi and PATH announce the launch of large-scale production of semisynthetic artemisinin against malaria. https://www.path.org/our-impact/media-center/sanofi-and-path-announce-the-launch-of-large-scale-production-of-semisynthetic-artemisinin-against-malaria/

Pedaling to the Top. (2025, January 23). Improve your mitochondria, improve your cycling [Video]. YouTube. https://www.youtube.com/watch?v=sy-4P_aN97k

Peplow, M. (2016). Synthetic biology’s first malaria drug meets market resistance. Nature, 530(7591), 389–390.

Piscisgate. (2014). Honeybee queen cells 14 [Photograph]. Wikimedia Commons, CC BY-SA 4.0. https://commons.wikimedia.org/wiki/File:Honeybee_queen_cells_14.jpg

Ro, D.-K., Paradise, E. M., Ouellet, M., Fisher, K. J., Newman, K. L., Ndungu, J. M., et al. (2006). Production of the antimalarial drug precursor artemisinic acid in engineered yeast. Nature, 440(7086), 940–943.

Russ London. (2010). Wellcome genome bookcase [Photograph]. Wikimedia Commons, CC BY-SA 3.0. https://commons.wikimedia.org/wiki/File:Wellcome_genome_bookcase.png

SciShow. (2015, July 6). How bees choose their queen [Video]. YouTube. https://www.youtube.com/watch?v=m_SlH3Uwslc

Senge, P. M. (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. Doubleday.

Spieth, J., & Lawson, D. (2006). Overview of gene structure. In WormBook. https://www.ncbi.nlm.nih.gov/books/NBK19701/

Stulp, G., Barrett, L., Tropf, F. C., & Mills, M. (2015). Does natural selection favour taller stature among the tallest people on earth? Proceedings of the Royal Society B, 282(1806), 20150211.

Tishkoff, S. A., Reed, F. A., Ranciaro, A., Voight, B. F., Babbitt, C. C., Silverman, J. S., et al. (2007). Convergent adaptation of human lactase persistence in Africa and Europe. Nature Genetics, 39(1), 31–40.

Visscher, P. M., Medland, S. E., Ferreira, M. A. R., Morley, K. I., Zhu, G., Cornes, B. K., et al. (2006). Assumption-free estimation of heritability from genome-wide identity-by-descent sharing between full siblings. PLoS Genetics, 2(3), e41.

World Health Organization. (2024). World malaria report 2024: Addressing inequity in the global malaria response. World Health Organization.

Yengo, L., Vedantam, S., Marouli, E., Sidorenko, J., Bartell, E., Sakaue, S., et al. (2022). A saturated map of common genetic variants associated with human height. Nature, 610(7933), 704–712.