How Inheritance Works at the Molecular Level

Four letters, read three at a time, and what comes out does the work

DNA is often called "the blueprint of life." Convenient, but it walks you into a ditch: a blueprint is a scaled drawing of the finished thing, and you can point at it and say "that is a window." Nowhere on DNA can you point and say "that is an eye."

What it actually is: a long chain strung with four kinds of small part, conventionally written A, T, C and G. The cell reads it three letters at a time, and each triplet stands for one amino acid. Amino acids link into a long chain, and the chain folds itself into a lump with a shape — that lump is a protein. Proteins are what actually do things: digesting milk, hauling oxygen around, making muscle contract. All proteins.

There is one step in between. The cell does not carry the original into the workshop; it first copies out a working duplicate (called RNA — much like DNA, but single-stranded and short-lived), sends the copy out to be used, and lets it rot afterwards. The step looks redundant and is in fact a control point: how many copies get made, and how long each survives, both change output while the original is untouched.

That table of "which triplet means which amino acid" was read out by brute force. In 1961 Nirenberg and Matthaei synthesised a stretch of RNA made of one letter repeated, dropped it into a pot of still-working juice squeezed out of bacteria, and got back protein chains built entirely from a single amino acid. That was the first line of the table. The rest was read out over the next few years, and the work took a Nobel Prize in 1968.

More important still: the table is nearly identical in bacteria, yeast, maize and you. Which is why, in 1978, people could put the human insulin gene into a gut bacterium and have the bacterium build human insulin; it reached the market in 1982. A bacterium can read a sentence written by a human — and only one explanation holds up: the table was already fixed in the common ancestor.

Nor is the table arranged at random. Similar triplets tend to stand for chemically similar amino acids, so a single miscopied letter often either changes no amino acid at all or swaps in one of much the same temperament. That is a layer of shock absorption. People have generated millions of random fake tables by computer for comparison, and the real one ranks very high for error tolerance.

But do not turn that into "evolution designed it to absorb errors" — nobody is designing. An error-tolerant table wastes fewer products, so lineages carrying it left more descendants; that is all. And a second explanation is on the table alongside. Crick suggested it may be a frozen accident: a table settled early and arbitrarily, which became unchangeable once every protein was built on top of it, because altering one letter would wreck all of them at once. Neither explanation has been ruled out, and most people think both are partly right.

Same text, and every cell reads only its own few passages

You have tens of trillions of cells, and with few exceptions every one of them holds the same DNA. The cell at the root of a hair and the cell in your pancreas that makes insulin hold the same text, word for word, yet they differ in shape, in lifespan, in what they do. The difference cannot be hiding in the text.

So where is it? Next to each gene — sometimes quite far away — sit short stretches that code for no protein at all, serving as docking points for another class of protein. When those proteins dock, they raise or lower how fast the neighbouring gene is copied out. Call these short stretches switches.

So a mutation has two routes to changing a body: alter the protein itself, or alter when, where, and how many copies get made. The second route is far cheaper — altering a protein means altering a tool that is in use everywhere, with consequences all over; altering a switch touches one part only.

Sticklebacks show this most clearly. Marine three-spined sticklebacks carry a pair of hard pelvic spines under the belly, like two small daggers, which lodge in a predator's mouth. When the glaciers retreated over ten thousand years ago, populations were stranded in newly formed inland lakes; the lakes held no such predators, but were short on calcium — and spines are expensive. Tens to hundreds of generations later, in many of those lakes, the spines are gone.

What matters is how they went. The Pitx1 gene itself is entirely intact, not a letter changed. What was lost is the neighbouring switch that works only in the pelvic region. Delete it, and Pitx1 stops being copied out there, so no spines grow; the same gene keeps working normally in the head and face. Researchers put the switch back, and the spines came back (Chan and colleagues, 2010). Better still, fish in different lakes lost it independently — and lost the same stretch.

The contrast makes it sharper: knock out the Pitx1 gene itself and, in lab animals, that is lethal. So changing a switch is not some secondary route — often it is the only route that goes anywhere.

Humans point the same way. Two decades of genome-wide association studies (scanning hundreds of thousands of genomes side by side against some trait or disease to see which positions vary along with it) have found that over ninety percent of associated positions sit outside protein-coding sequence. One honest caveat: association is not causation, and landing in non-coding sequence does not automatically make a position "the switch" — going from an associated position to an actual mechanism often takes years. But the direction is clear: the common differences between people mostly touch the reading, not the text.

Cells also remember who they are: a liver cell divides into liver cells, carrying that "which passages to read" state along with it. Whether such marks can cross the sperm and egg into the next generation is a separate question, and most are wiped and reset around fertilisation.

The central dogma says you cannot go back, not that you cannot go around

Nearly everyone has heard it: the central dogma is DNA to RNA to protein, a one-way street. Then they hear that retroviruses copy RNA back into DNA, and conclude that "the central dogma has been overturned."

That is not what Crick wrote in 1958. He said: once sequence information has passed into a protein, it cannot get out again — no mechanism reads a protein and works backwards to write out a stretch of nucleic acid. Nothing has touched that claim to this day. What was overturned was the slogan version.

The detours are real enough. Reverse transcription: in 1970 Temin and Baltimore independently found the enzyme that copies RNA into DNA, the one HIV depends on. Traces of that are in you — roughly 8% of the human genome is the wreckage of ancient viral sequence, and close to half is piled up from stretches that copy themselves and insert elsewhere. Information is still moving between nucleic acids, so Crick's actual claim is untouched; the popular version collapsed.

RNA is not only a middleman: the machine that strings amino acids together is called the ribosome, and the core that does the real chemistry is RNA, not protein (structures solved around 2000, Nobel Prize in Chemistry in 2009).

One gene, one protein? Also wrong: one copy can have different passages cut out and be spliced back together into several different proteins. Before sequencing, the common guess was a hundred thousand human genes; the count came back at roughly twenty thousand protein-coding genes, not far off a roundworm. Much of the extra complexity lives in the splicing and the reading.

Prions are the closest thing to protein carrying information: a misfolded protein forces its fellows into the same wrong fold, on and on. But what travels is a shape, not a sequence, so on Crick's exact words it still does not violate anything — how best to phrase that point is not fully settled.

Why do all these "exceptions" persist? Because selection does not audit rules for tidiness, only copy number. Reverse transcription is useful to a virus, so it stayed. The self-copying stretches occupying nearly half the genome need not benefit the host at all; they spread because they copy themselves fast, which is a different ledger entirely from "is this good for the animal" (the gene's-eye-view topic is about exactly how those two ledgers are kept apart). That clean arrow in the textbook was drawn to make teaching easier. Inside the cell there was never any such line.