When a double helix of DNA is replicated, each of the two original strands serves as a template for building a new complementary strand. The result is two identical double helices, each containing one old strand and one newly built strand. This process, called semiconservative replication, was confirmed experimentally in 1958 by Matthew Meselson and Franklin Stahl using heavy and light nitrogen isotopes to track which strands ended up in daughter molecules. In human cells, the entire genome is copied during a roughly 8 to 10 hour window within the cell cycle known as S phase.
How the Double Helix Unwinds
Before any copying can begin, the two strands of the helix must be separated. An enzyme called helicase burns through ATP to travel along the DNA and pry the strands apart, typically starting in regions rich in A-T base pairs (which are held together by two hydrogen bonds instead of three, making them easier to split). This creates a Y-shaped structure called a replication fork, where the actual copying machinery assembles.
Unwinding the helix creates two problems. First, the exposed single strands are fragile and tend to fold back on themselves, so proteins called single-strand binding proteins coat them to keep them stable and accessible. Second, separating the strands ahead of the fork causes the DNA further down the line to twist tighter, like winding a rubber band. Enzymes called topoisomerases relieve this tension by temporarily cutting one or both strands, letting the helix rotate, and then resealing the break.
Building the New Strands
The enzyme that actually assembles new DNA is DNA polymerase. It reads the template strand and adds matching nucleotides one at a time, linking each new building block to the growing chain through a chemical bond called a phosphodiester bond. The energy for each addition comes from the incoming nucleotide itself: it arrives carrying three phosphate groups, and when two of those are clipped off, the released energy (about 6.2 kilocalories per nucleotide) drives the new bond into place.
There’s a catch. DNA polymerase can only add nucleotides in one direction (5′ to 3′) and it can’t start from scratch. It needs a short RNA primer, about 10 nucleotides long, laid down by an enzyme called primase. This gives the polymerase a starting point with the right chemical handle (a free 3′-OH group) to begin extending.
Leading Strand vs. Lagging Strand
Because the two strands of DNA run in opposite directions and polymerase only works in one direction, the two new strands are built differently. One strand, called the leading strand, points in the same direction the replication fork is moving. It only needs a single primer and then polymerase can follow the fork smoothly, synthesizing a continuous stretch of new DNA.
The other strand, the lagging strand, runs the opposite way. Polymerase has to work in short bursts, building small fragments of about 100 to 200 nucleotides each, known as Okazaki fragments. Each fragment needs its own RNA primer. As one fragment is completed, a new primer is laid down closer to the fork and the next fragment begins. This means the lagging strand requires many more initiation events and a more complex cleanup process afterward.
Joining the Pieces Together
Once the Okazaki fragments are built, the RNA primers scattered throughout the lagging strand must be removed and replaced with DNA. Specialized enzymes cut out the RNA and a different polymerase fills the resulting gaps using the parent strand as a guide. Finally, an enzyme called DNA ligase seals the remaining nicks by forming the last phosphodiester bonds between adjacent fragments. The end result is a smooth, continuous new strand on both sides of the fork.
Error Correction and Accuracy
DNA polymerase is remarkably precise, but it still makes mistakes, roughly one error for every 10,000 to 100,000 nucleotides it adds. To catch these, polymerase has a built-in proofreading function: it can detect a mismatched base, reverse direction, remove the wrong nucleotide, and try again. This proofreading step improves accuracy by 10- to 100-fold in lab conditions, and potentially much more inside a living cell.
After replication is complete, a second safety net called mismatch repair scans the newly made DNA for errors that proofreading missed. Together, these systems bring the final error rate in normal cells down to approximately 0.021 mutations per 100 million nucleotides per generation. For a human genome of over 6 billion base pairs, that translates to only a handful of uncorrected mistakes per cell division.
The Problem at Chromosome Ends
Linear chromosomes face a unique challenge that circular bacterial DNA does not. When the final RNA primer on the lagging strand is removed at the very tip of a chromosome, there’s no upstream DNA for polymerase to extend from, leaving a small gap. This means chromosomes lose about 50 to 150 base pairs from their ends with every round of replication.
To buffer against this, the ends of chromosomes are capped with repetitive sequences called telomeres (TTAGGG, repeated thousands of times in humans). Telomeres don’t carry essential genetic information. They exist to absorb the loss. In stem cells and reproductive cells, an enzyme called telomerase can extend telomeres by adding new repeats, using a built-in RNA template as a guide. Most other cells in the body lack significant telomerase activity, which is why telomeres gradually shorten with age. Cancer cells, notably, often reactivate telomerase to maintain their telomeres indefinitely, which is one reason they can keep dividing without limit.
The very tip of each chromosome also folds into a protective loop structure, preventing the cell’s repair machinery from mistaking the natural chromosome end for a dangerous break in the DNA.
How Long Replication Takes
Speed varies dramatically between organisms. Bacterial cells, which have a single circular chromosome, can replicate their entire genome in under an hour using just two replication forks moving in opposite directions from a single starting point. Human cells face a much bigger task: 6 billion base pairs spread across 46 chromosomes. To finish within the 8 to 10 hour S phase window, cells fire up thousands of replication origins across the genome simultaneously, with each origin generating two forks that move outward in both directions. When neighboring forks meet, their newly synthesized stretches are joined together by ligase, the same way Okazaki fragments are connected on the lagging strand.
Cells don’t activate all origins at once. Different regions of the genome replicate at different times during S phase, following a consistent program. If the cell detects DNA damage or stalled forks, a checkpoint mechanism can pause the firing of later origins, extending S phase to allow time for repair before replication continues.