Knowing that genes are DNA raised the next question: what does a gene actually do? Part of the answer had come early. In 1902 the English physician Archibald Garrod studied alkaptonuria, a rare condition in which urine darkens on standing. He noticed it ran in families, often among children of cousin marriages, and argued that it was a recessive Mendelian trait caused by a missing step in the body's chemistry. He later called such conditions inborn errors of metabolism. Few people took notice for decades.

One gene, one enzyme

In 1941 George Beadle and Edward Tatum at Stanford exposed the bread mold Neurospora to X-rays to create mutants. Some mutants could grow only if a particular nutrient was added to their food, because they could no longer make it themselves. Each such mutation knocked out a single step in a chemical pathway. Beadle and Tatum concluded that each gene controls the making of one enzyme. The idea was later refined, since many proteins are not enzymes and some are assembled from several gene products, but it established that genes work by directing the making of proteins.

The central dogma

Proteins do most of the work in a cell: they speed up reactions, form structures, carry oxygen and send signals. A protein is a chain of amino acids folded into a shape, and its sequence determines that shape. In 1958 Francis Crick set out what he called the central dogma: information flows from DNA to RNA to protein, and once information has passed into protein it cannot flow back out. In the first step, transcription, a gene's DNA is copied into messenger RNA, a similar single-stranded molecule in which the base uracil (U) replaces thymine. In the second, translation, cell machines called ribosomes read the messenger RNA and string amino acids together in the order it specifies. Later discoveries added detours, such as viruses like HIV that copy RNA back into DNA, but no flow from protein back to genes has ever been found.

Cracking the code

Four bases must specify twenty amino acids. Pairs of bases give only sixteen combinations, too few; triplets give sixty-four, more than enough. Experiments confirmed that the code is read in triplets of bases called codons. In 1961 Marshall Nirenberg and Heinrich Matthaei made an artificial RNA containing only uracil and added it to a cell extract. It produced a protein chain made only of phenylalanine, so UUU, coding for phenylalanine, was the first codon deciphered. By 1966 all sixty-four were known. Sixty-one specify amino acids and three are stop signals, so the code is redundant: most amino acids have several codons. It is also nearly universal: almost all living things read the same code. That is why a human insulin gene placed in bacteria can make human insulin, as it has commercially since 1982.

Genes in pieces

In 1977 teams led by Phillip Sharp and Richard Roberts found a surprise. In many organisms, including humans, genes are interrupted by stretches that do not code for protein, called introns. After transcription the introns are cut out and the remaining pieces, called exons, are spliced together. Cells can splice the same gene in different ways, a process called alternative splicing, which lets one gene make several different proteins.

Switching genes on and off

Almost every cell in your body carries the same DNA, yet a nerve cell and a skin cell look and act nothing alike. The difference is which genes are switched on. The first switch was worked out in bacteria by François Jacob and Jacques Monod, published in 1961. The gut bacterium E. coli makes the enzymes to digest the milk sugar lactose only when lactose is present. A repressor protein normally sits on the DNA and blocks those genes; lactose pulls it off. This system, the lac operon, showed that genes can be switched on and off as conditions change. In complex organisms regulation involves many layers, and much of the DNA outside genes consists of switches that control when and where genes act.

Worked example: reading a message

Take a short stretch of messenger RNA: AUG GAG UUU UAA. A ribosome starts at AUG, which codes for methionine and also serves as the usual start signal. GAG codes for glutamic acid, UUU for phenylalanine, and UAA is a stop codon, so the result is a chain of three amino acids. Now consider single-letter changes. If GAG becomes GAA, the protein is unchanged, because GAA also codes for glutamic acid. That is a silent change, possible because the code is redundant. If GAG becomes GUG, glutamic acid is replaced by valine, a missense change. This exact swap, in a gene for part of hemoglobin, causes sickle cell disease. If UUU becomes UAA, the chain stops early, a nonsense change that usually wrecks the protein. And if a single letter is deleted, every codon after it is misread, a frameshift. Those four outcomes are the vocabulary of the next level.