Molecular Basis of Inheritance

Any molecule that acts as the genetic material has to pass four tests. It must be able to make an exact copy of itself. It must be chemically and structurally stable. It must allow slow changes, that is mutations, so that evolution is possible. And it must be able to express itself as the characters an organism actually shows. For a long time proteins were the favourite candidate, because proteins are built from twenty different amino acids and looked complex enough to store information, while DNA was dismissed as a dull repeat of only four bases. Three experiments settled the question in favour of DNA.

Frederick Griffith worked in 1928 with Streptococcus pneumoniae, a bacterium that causes pneumonia. It grows as two strains. The S strain forms smooth, shiny colonies because every cell is wrapped in a polysaccharide capsule, and it is virulent, so injecting it kills mice. The R strain forms rough colonies, has no capsule, and does not kill mice. Griffith found that heat-killed S bacteria on their own were harmless. But when he injected heat-killed S bacteria together with live R bacteria, the mice died, and living S bacteria could be recovered from their bodies. Something released from the dead S cells had permanently changed harmless R cells into capsule-making S cells, and the change was inherited by their offspring. Griffith called this something the transforming principle and the process transformation. He could not say what the chemical was.

Oswald Avery, Colin MacLeod and Maclyn McCarty spent years purifying that chemical from heat-killed S cells. Their method was to destroy one class of molecule at a time and then check whether transformation still happened. Adding proteases, which digest proteins, did not stop transformation. Adding RNase, which digests RNA, did not stop it either. Adding DNase, which digests DNA, stopped transformation completely. The conclusion was direct: DNA is the transforming principle, and therefore DNA is the hereditary material. Many biologists still hesitated, partly because a purified DNA preparation could carry traces of protein along with it.

Alfred Hershey and Martha Chase removed the last doubt in 1952 using bacteriophage T2, a virus that infects Escherichia coli. A phage is made of only two things, a protein coat and DNA inside it, and the two differ in one very useful way. DNA contains phosphorus but no sulphur, while protein contains sulphur in the amino acids cysteine and methionine but no phosphorus. So they grew one batch of phages in a medium containing radioactive phosphorus, which labelled only the phage DNA, and a second batch in a medium containing radioactive sulphur, which labelled only the phage protein. Each batch was allowed to infect bacteria. The cultures were then whirled in a blender, which shook the empty viral coats off the bacterial surface, and spun in a centrifuge, which pulled the heavier bacterial cells to the bottom. Bacteria infected by the phosphorus-labelled phages were radioactive. Bacteria infected by the sulphur-labelled phages were not. Only DNA had entered the bacterial cell, so DNA and not protein carries the instructions for building new viruses.

DNA won the argument for a chemical reason too. RNA has a hydroxyl group on the 2' carbon of every sugar, and that reactive group makes RNA easy to break down. RNA is also catalytic, and a reactive molecule is an unstable one. DNA lacks that 2' hydroxyl, uses thymine in place of uracil, which adds further stability, and is double stranded, so a damaged strand can be repaired using the intact partner as a reference. RNA still serves as the genetic material in some viruses, and such viruses mutate much faster. There is one test, though, on which RNA does better than DNA. RNA can be translated into protein directly, whereas DNA has to be transcribed into RNA first before any protein can be made from it. So RNA expresses itself more readily, while DNA is the safer long-term store. That is the division of labour we actually see in cells: DNA keeps the information and RNA carries and uses it.

A nucleotide has three parts: a nitrogenous base, a pentose sugar and a phosphate group. The sugar is deoxyribose in DNA and ribose in RNA. A base joined to the sugar alone is a nucleoside; add the phosphate and it becomes a nucleotide. The bases come in two chemical types. Purines, which are adenine and guanine, have two fused rings. Pyrimidines, which are cytosine and thymine in DNA and cytosine and uracil in RNA, have a single ring. Nucleotides are linked into a chain by phosphodiester bonds, each joining the 3' carbon of one sugar to the 5' carbon of the next through a phosphate. This gives every strand a direction, called polarity: one end has a free 5' phosphate and the other a free 3' hydroxyl.

Erwin Chargaff analysed DNA from many organisms and found a rule hiding in the numbers. In any double-stranded DNA the amount of adenine always equals the amount of thymine, and the amount of guanine always equals the amount of cytosine. It follows that total purines equal total pyrimidines, so the ratio (A+G)/(T+C) works out to 1. Note that the ratio (A+T)/(G+C) is not fixed; it changes from one organism to another, and that is exactly why it can be used to tell species apart.

In 1953 James Watson and Francis Crick, using the X-ray diffraction photographs produced by Maurice Wilkins and Rosalind Franklin, proposed the double helix. Two polynucleotide chains wind around a common axis in a right-handed spiral. The sugar-phosphate backbones lie on the outside and the bases point inward, stacked nearly flat one above the other like the steps of a spiral staircase. The two chains are antiparallel: where one runs 5' to 3', its partner alongside runs 3' to 5'. The chains are held together by hydrogen bonds between the bases, and the pairing is fixed. Adenine pairs with thymine through two hydrogen bonds, and guanine pairs with cytosine through three. Because a two-ring purine always faces a one-ring pyrimidine, the helix keeps a uniform width of about 2 nm all along its length. Successive base pairs sit 0.34 nm apart, there are roughly ten base pairs in one complete turn, and so one turn, called the pitch, measures about 3.4 nm. DNA rich in G and C needs more heat to separate, because each of those pairs is held by an extra hydrogen bond.

Now the packing problem. A human cell carries about 6.6 x 10⁹ base pairs of DNA, and at 0.34 nm per base pair that stretches to roughly 2.2 metres of thread. It has to fit inside a nucleus only a few micrometres across. The solution is to wind the DNA on protein spools. DNA is negatively charged because of its phosphate groups, and histones are basic proteins carrying a positive charge, because they are rich in the positively charged amino acids lysine and arginine, so the two stick to each other. Eight histone molecules, two copies each of H2A, H2B, H3 and H4, form a histone octamer. About 200 base pairs of DNA wrap around one octamer, and the DNA-plus-octamer unit is called a nucleosome. Histone H1 sits on the outside and seals the DNA where it enters and leaves the bead. Under an electron microscope a gently stretched chromatin fibre looks like beads threaded on a string, each bead a nucleosome and the string the linker DNA joining them. These nucleosomes then coil and fold on themselves repeatedly, helped by a set of additional proteins called non-histone chromosomal proteins, until the whole length fits. In an interphase nucleus the loosely packed, lightly stained regions are euchromatin and are transcriptionally active, while the tightly packed, darkly stained regions are heterochromatin and are inactive.

Watson and Crick pointed out in their own paper that the fixed base pairing suggests an obvious copying mechanism. If the two strands separate, each one specifies the sequence of its new partner. This predicts that every daughter DNA molecule will contain one old strand and one newly made strand, which is what semi-conservative replication means. Two rival ideas had to be ruled out: conservative replication, in which the original molecule stays whole and a completely new molecule is made alongside it, and dispersive replication, in which old and new stretches are mixed through both strands.

Matthew Meselson and Franklin Stahl tested this in 1958. They grew Escherichia coli for many generations in a medium where the only nitrogen source was ammonium chloride made with the heavy isotope of nitrogen, so every nitrogen atom in the bases was heavy and the whole DNA was denser than normal. The cells were then shifted to a medium containing ordinary light nitrogen, and samples were taken at regular intervals. The DNA from each sample was spun in a caesium chloride density gradient, a technique that lets each molecule settle at the level matching its own density, so molecules of different weight form separate bands. E. coli divides about every twenty minutes in these conditions. After twenty minutes, that is one round of replication, all the DNA formed a single band of intermediate density, exactly halfway between heavy and light. That result alone kills conservative replication, which would have produced one heavy band and one light band with nothing in between. After forty minutes, two rounds, there were two bands in equal amounts, one hybrid and one light. That pattern is what semi-conservative replication predicts and dispersive replication does not. Similar experiments by Taylor and his colleagues on root tips of Vicia faba, the broad bean, using radioactive thymidine, showed that whole chromosomes replicate the same way.

Replication does not begin just anywhere. It starts at a defined sequence called the origin of replication. A bacterial chromosome has a single origin; a long eukaryotic chromosome has many, so the copying can be shared out. Helicase unwinds the double helix, single-strand binding proteins hold the separated strands apart so they do not snap back together, and topoisomerase (DNA gyrase in bacteria) releases the twisting strain that builds up ahead of the unwinding point. The Y-shaped junction where the two strands part is the replication fork.

The central enzyme is DNA-dependent DNA polymerase. Its raw materials are deoxyribonucleoside triphosphates, and these do double duty: they supply the nucleotide and, when their two terminal phosphates are cut off, they supply the energy for joining it on. The enzyme has two firm limitations that shape the whole process. First, it can add a nucleotide only to a free 3' hydroxyl group, so a new strand can grow only in the 5' to 3' direction. Second, it cannot start a chain on bare template, so a short RNA primer laid down by primase provides the first 3' end for it to build from.

Because the two template strands run in opposite directions, only one new strand can be built continuously in the same direction that the fork is opening. That one is called the leading strand. On the other template the enzyme has to work away from the fork, so the new strand is built in short pieces called Okazaki fragments, each started fresh as more template is exposed; this is the lagging strand and its synthesis is discontinuous. DNA ligase then seals the nicks between the fragments to give one continuous strand. Notice that ligase does not make new strand, it only joins ends. Speed matters here: the E. coli genome is about 4.6 x 10⁶ base pairs, and it is copied at roughly 2000 base pairs per second. Accuracy matters just as much, because every uncorrected mistake becomes a mutation. In a eukaryotic cell, replication is confined to the S phase of interphase, and it is tightly linked to cell division; if the DNA replicates but the cell then fails to divide, the chromosome number doubles and polyploidy results.

Before the details, the shape of the whole process. Francis Crick set out the direction in which genetic information moves inside a cell as DNA to RNA to protein, and this statement is called the central dogma. DNA is copied into RNA by transcription, and the RNA is then read into a chain of amino acids by translation. In some viruses the first arrow runs backwards, so that RNA is copied into DNA, but for the rest of life the order is fixed. This section deals with the first arrow.

Transcription is the copying of one segment of DNA into RNA. It differs from replication in two ways: only a small part of the DNA is copied, and only one of the two strands is used as the template. There is a good reason for using only one strand. If both were transcribed, they would produce two different RNA molecules coding for two different proteins, which would make one gene mean two things; worse, the two RNAs would be complementary to each other, so they would pair up into a double-stranded molecule and translation would stop.

The stretch of DNA that is transcribed as one piece is called a transcription unit, and it has three parts. The promoter comes first and is the landing site for RNA polymerase; it lies towards the 5' end of the coding strand, which is why it is described as upstream. The structural gene comes next. The terminator comes last, towards the 3' end of the coding strand, and it is where transcription stops.

Of the two DNA strands, the one actually read by the enzyme is the template strand, and it runs 3' to 5' in the direction the polymerase travels. The other strand is called the coding strand. It is never copied, yet it is the more useful one to look at, because its sequence is identical to the RNA that is made, with thymine replaced by uracil everywhere. So if you are given the coding strand and asked for the mRNA, you simply rewrite it and change every T to a U. RNA polymerase, like DNA polymerase, can only build in the 5' to 3' direction, and that is exactly why it has to read the template from the template's 3' end towards its 5' end.

In prokaryotes a single RNA polymerase makes all three kinds of RNA, that is mRNA, tRNA and rRNA. It cannot start on its own: an accessory protein called the sigma factor associates with it for initiation and leaves once transcription is under way. Termination needs another accessory protein, the rho factor. Because a bacterial cell has no nuclear membrane, ribosomes can begin translating an mRNA while its far end is still being transcribed, so transcription and translation are coupled in time and place. Bacterial mRNA is often polycistronic, meaning one mRNA molecule carries several genes read one after another, and it needs no processing before use.

In eukaryotes the work is divided among three nuclear RNA polymerases. RNA polymerase I transcribes the ribosomal RNAs 28S, 18S and 5.8S. RNA polymerase II transcribes the precursor of mRNA, called heterogeneous nuclear RNA or hnRNA. RNA polymerase III transcribes tRNA, the 5S rRNA and the small nuclear RNAs. Eukaryotic mRNA is monocistronic, that is one mRNA for one polypeptide.

Eukaryotic genes are also split genes. The coding stretches, called exons, are interrupted by non-coding stretches called introns, whose other name, intervening sequences, says exactly what they are. Only exons survive into the mature message. The primary transcript therefore has to be processed in three ways inside the nucleus. Splicing removes the introns and joins the exons together in their original order. Capping adds an unusual nucleotide, methyl guanosine triphosphate, to the 5' end. Tailing adds a stretch of about 200 to 300 adenylate residues to the 3' end, and this poly-A tail is put on without any template to copy from. Only after capping, tailing and splicing does the hnRNA become mature mRNA and move out of the nucleus to the ribosomes. Because different combinations of exons can sometimes be joined, one split gene can yield more than one protein, which is part of why humans have far fewer genes than proteins.

The genetic code is the rule that converts a base sequence into an amino acid sequence. The size of the code word can be worked out by counting. There are four bases. If one base named one amino acid, only four amino acids could be specified. If two bases were read together, there would be four times four, that is sixteen combinations, still short of the twenty amino acids that proteins use. Three bases give four times four times four, which is sixty-four, comfortably more than enough. George Gamow argued on exactly these mathematical grounds that the code must be a triplet code. It was established experimentally through Marshall Nirenberg's cell-free protein synthesising system and Har Gobind Khorana's chemically synthesised RNA molecules with defined repeating sequences, helped by the enzyme discovered by Severo Ochoa, which could polymerise RNA of known composition.

Of the sixty-four codons, sixty-one specify amino acids. The remaining three, UAA, UAG and UGA, do not code for any amino acid at all and act as stop or termination signals. AUG has two jobs at once: it is the initiation codon, and it also codes for the amino acid methionine wherever it appears inside a message. The code is degenerate, which means that most amino acids are specified by more than one codon. Degenerate does not mean sloppy: the code is also unambiguous and specific, because any one codon codes for one amino acid and no other. It is read continuously with no punctuation between successive codons, which is what comma-less means, and it is non-overlapping, so each base belongs to exactly one codon. Finally it is nearly universal: the same codon means the same amino acid in a bacterium and in a human being, with a small number of exceptions such as the code used inside mitochondria.

Because the code has no commas, the reading frame is fixed entirely by where translation starts. That has a consequence worth understanding. If one or two bases are inserted into or deleted from a gene, every codon after that point is read in the wrong grouping, and the rest of the protein comes out as nonsense. This is a frameshift mutation. If three bases are added or lost together, the frame is restored after that point, so only one amino acid is gained or lost and the rest of the protein is normal. A different kind of change is the substitution of a single base, which alters only one codon. In sickle cell anaemia, a single base substitution changes the sixth codon of the gene for the beta chain of haemoglobin from GAG to GUG, so glutamic acid at the sixth position of that chain is replaced by valine, and that one swap changes the behaviour of the whole haemoglobin molecule.

There is a gap in this story that has to be filled. A codon is a sequence of bases and an amino acid is a completely different sort of molecule; the two cannot recognise each other directly. Francis Crick predicted that some adapter molecule must exist with two ends, one that reads the code and one that carries the amino acid. That molecule turned out to be transfer RNA. Each tRNA has an anticodon loop carrying three bases complementary to the codon it must recognise, and an amino acid acceptor end at which the correct amino acid is attached. There is a separate tRNA for each amino acid, plus a special initiator tRNA for the start codon. Importantly, there are no tRNAs for the three stop codons, which is precisely why the ribosome cannot go past them. Drawn flat on paper, a tRNA looks like a clover leaf, but in the cell it folds up into a compact shape resembling an inverted L, with the anticodon at one tip and the amino acid at the other.

Translation is the building of a polypeptide from the order of bases in an mRNA. The first step happens before any ribosome is involved. Each amino acid is joined to its own tRNA in a reaction that consumes ATP, a step called activation of the amino acid or charging of the tRNA, also known as aminoacylation. Only a charged tRNA can be used, and once two charged tRNAs are held side by side on the ribosome, a peptide bond can form between the amino acids they carry.

The ribosome is the factory where this happens. It is made of two subunits which lie apart from each other when the ribosome is inactive; when the smaller subunit meets an mRNA the two subunits come together and translation begins. A bacterial ribosome is a 70S particle made of a 50S and a 30S subunit, while a eukaryotic cytoplasmic ribosome is 80S, made of a 60S and a 40S subunit. The larger subunit has grooves that can hold two tRNAs next to each other, so that the growing chain on one can be transferred onto the amino acid on the next. The peptide bond is formed not by a protein enzyme but by one of the ribosomal RNA molecules acting as a catalyst, which is why it is called a ribozyme.

An mRNA carries more than just the coding message. Between its 5' end and the start codon lies a stretch that is never translated, and a second untranslated stretch follows the stop codon before the 3' end. These untranslated regions carry no amino acid information but are needed for translation to run efficiently. Translation itself has three phases. In initiation, the ribosome assembles on the mRNA at the start codon and only the initiator tRNA is accepted there. In elongation, charged tRNAs arrive one after another, each anticodon pairing with the next codon, and the polypeptide grows by one amino acid per codon while the ribosome moves along. In termination, a stop codon arrives, no tRNA can pair with it, a release factor binds instead, the finished polypeptide is set free and the two subunits separate.

A cell does not need every protein all the time, and making an enzyme for a food that is not present is pure waste. In bacteria the main control point is transcription, and the classic example is the lac operon, worked out by Francois Jacob and Jacques Monod. An operon is a set of structural genes transcribed together as a single unit, together with the switches placed in front of them. Along the DNA, in order, the lac system has the regulatory gene i, then the promoter, then the operator, and then three structural genes named z, y and a in that sequence. Gene z codes for beta-galactosidase, the enzyme that splits lactose into galactose and glucose. Gene y codes for permease, which increases the permeability of the cell to lactose. Gene a codes for transacetylase. All three are transcribed together into one long polycistronic mRNA, so they are made and stopped as a group.

The i gene is transcribed at a low level all the time and produces a protein called the repressor. When lactose is not available, the repressor binds tightly to the operator, which sits between the promoter and gene z. RNA polymerase can still settle on the promoter but cannot move forward past the occupied operator, so no mRNA for the three enzymes is produced. When lactose is supplied, a small amount enters through the few permease molecules the cell always carries, and lactose then acts as the inducer. It binds to the repressor and changes its shape so that the repressor can no longer hold on to the operator. The operator is now free, RNA polymerase moves through, the polycistronic mRNA is made and the three enzymes appear. Because the substrate itself turns the genes on, the lac operon is described as an inducible operon, and because the controlling protein is a repressor whose job is to block transcription, this arrangement is called negative regulation. When the lactose has all been used up, there is nothing left to hold the repressor in its inactive shape, it binds the operator again, and the operon shuts down.

The Human Genome Project ran from 1990 to 2003. In the United States it was coordinated by the Department of Energy and the National Institutes of Health, and the Wellcome Trust in Britain joined as a major partner. The goals were to determine the order of all roughly 3 x 10⁹ base pairs in a human genome, to identify the genes within it, to store the information in databases and to improve the tools for analysing such data. It was called a megaproject for good reason. At a sequencing cost of about three US dollars per base, the total estimate came to about nine billion US dollars, and the sequence printed as books of a thousand pages each would fill a shelf 3300 volumes long. Because the information touches individuals so directly, the ethical, legal and social questions it raises were made part of the project from the start.

Two broad approaches were used. The first identified only the genes that are actually expressed as RNA, using short sequences read from them called Expressed Sequence Tags. The second, called sequence annotation, sequenced the whole genome, coding and non-coding regions alike, and only afterwards assigned functions to the different parts. The practical procedure was to break the DNA into small fragments, insert each fragment into a vector so that it could be multiplied inside a host cell, sequence the fragments in automated sequencers based on the method developed by Frederick Sanger, and then use computer programs to spot overlaps and assemble the fragments back into a continuous sequence. The vectors used for the very large pieces were bacterial artificial chromosomes and yeast artificial chromosomes. The assembled sequences were then assigned to particular chromosomes using genetic and physical maps built from polymorphism in restriction enzyme recognition sites and from repetitive microsatellite sequences.

The findings were full of surprises. The human genome contains about 3164.7 million base pairs. An average gene is roughly 3000 bases long, but the largest known human gene, dystrophin, stretches to 2.4 million bases. The number of genes came out at about 30000, far below earlier estimates of 80000 to 140000, and less than 2 per cent of the genome actually codes for proteins, with repeated sequences making up a very large part of the remainder. Chromosome 1 carries the most genes, 2968 of them, and the Y chromosome the fewest, 231. About 99.9 per cent of the bases are exactly the same in every human being, and around 1.4 million sites were found where individuals differ by a single base; these are called single nucleotide polymorphisms, or SNPs, and they are what makes it possible to track down chromosomal regions linked to disease. The function of more than half the genes discovered is still not known. Alongside the human genome, the genomes of several model organisms were sequenced, including rice, Arabidopsis thaliana, the fruit fly Drosophila melanogaster and the free-living non-pathogenic nematode Caenorhabditis elegans.

DNA fingerprinting, developed by Alec Jeffreys, works on the small fraction of the genome that differs from person to person, and most of that variation lies in non-coding repetitive DNA. When total genomic DNA is spun in a density gradient, these repeated stretches separate out as small peaks beside the main band, which is why they are called satellite DNA. They are classified by the length of the repeating unit into microsatellites and minisatellites. A particular class of minisatellite is the VNTR, short for variable number of tandem repeats. Its repeat unit is short, but the number of copies varies enormously from one individual to another, so the fragment produced can be anywhere from roughly 0.1 to 20 kilobases long. A person inherits one set of these repeats from each parent, so the overall pattern is effectively unique except in identical twins, and it is the same in DNA taken from any tissue of that person, whether blood, hair follicles, skin, bone, saliva or semen.

The technique runs in a fixed order. DNA is isolated from the sample. It is cut into fragments using restriction endonucleases. The fragments are separated according to size by gel electrophoresis. The separated fragments are transferred, or blotted, from the gel onto a synthetic membrane of nitrocellulose or nylon, a step named Southern blotting after its inventor. The membrane is then treated with a labelled VNTR probe, which sticks only to fragments containing that repeat, a process called hybridisation. Finally the positions where the probe has bound are revealed by autoradiography. The result is a pattern of bands of different sizes, and that pattern is characteristic of the individual. It is used in forensic investigation, in resolving questions of parentage, and in studying genetic diversity within and between populations.