DNA polymerase is the enzyme that connects new nucleotides together during DNA replication and proofreads them for errors. It does both jobs using two separate active sites built into the same protein, one for building the new DNA strand and another for removing mistakes.
How DNA Polymerase Connects Nucleotides
DNA polymerase works by adding one nucleotide at a time to the growing end of a new DNA strand. Each time it adds a nucleotide, it forms a phosphodiester bond, which is the chemical link between the sugar of one nucleotide and the phosphate group of the next. This creates the sugar-phosphate backbone that holds the DNA strand together.
The enzyme always builds in one direction: from the 5′ end to the 3′ end of the new strand. Think of it like a zipper that can only close in one direction. At each step, DNA polymerase reads the template strand (the existing strand being copied), selects the matching nucleotide from the surrounding pool, and snaps it into place. Adenine pairs with thymine, and cytosine pairs with guanine.
One important quirk: DNA polymerase cannot start a new strand from scratch. It needs a short starter piece called a primer, which provides the free 3′ end that the enzyme attaches the first new nucleotide to. A different enzyme (primase) lays down this short RNA primer, and DNA polymerase takes over from there.
How Proofreading Works
DNA polymerase makes mistakes roughly once every 10,000 to 100,000 nucleotides it adds. That sounds rare, but the human genome has over 6 billion base pairs, so without error correction, every cell division would introduce tens of thousands of mutations. Proofreading catches most of those mistakes before they become permanent.
The enzyme’s proofreading ability comes from a built-in 3′ to 5′ exonuclease activity, meaning it can chew backward along the strand it just built. When DNA polymerase inserts the wrong nucleotide, the mismatched base pair fits poorly in the enzyme’s active site. This triggers the enzyme to pause, shift the strand tip to its separate proofreading site (located about 35 angstroms away from the building site), and clip out the incorrect nucleotide. Then the strand snaps back into the building site, and the enzyme tries again with the correct match. A mismatched base pair at the tip of the growing strand is the preferred target for this exonuclease, so the enzyme is much more likely to remove a wrong base than a correct one.
The Enzyme’s Physical Structure
The shape of DNA polymerase is often compared to a right hand. The “palm” region contains the active site where new nucleotides are bonded to the growing strand. The “fingers” help grip the incoming nucleotide and the template strand, positioning everything correctly. The “thumb” wraps around the newly formed double-stranded DNA to keep the enzyme from sliding off. The cleft formed by these three regions is roughly 20 to 24 angstroms wide and 25 to 35 angstroms deep.
The proofreading exonuclease domain sits as a separate region at one end of the protein. When a mismatch is detected, the end of the growing strand physically moves from the palm’s building site over to this exonuclease domain, where the bad nucleotide is removed.
Different Versions in Different Organisms
Both bacteria and human cells use DNA polymerase for replication, but they use different versions of the enzyme. In bacteria like E. coli, the main replicating enzyme is DNA polymerase III. In human and other eukaryotic cells, two versions share the workload: DNA polymerase epsilon handles the leading strand (the strand built continuously), while DNA polymerase delta handles the lagging strand (the strand built in short fragments). Both of these eukaryotic polymerases have proofreading ability.
Interestingly, DNA polymerase delta can also proofread errors made by DNA polymerase epsilon. Research has shown that when polymerase epsilon makes more mistakes than usual, polymerase delta catches and corrects many of those errors on the leading strand, acting as a backup proofreader.
Backup Systems Beyond Proofreading
Even with proofreading, DNA polymerase still lets some errors slip through. A second layer of defense called mismatch repair scans the newly built DNA after replication is complete. Specialized proteins detect spots where the bases don’t pair correctly, cut out the section containing the error, and let DNA polymerase fill in the gap with the correct sequence.
Together, these systems are remarkably effective. DNA polymerase’s initial nucleotide selection produces about 1 error per 10,000 to 100,000 bases. Proofreading removes most of those. Mismatch repair catches most of what’s left. The final result is a spontaneous mutation rate of roughly 1 error per 10 billion base pairs per cell division. That means your cells copy 6 billion letters of DNA with, on average, fewer than one uncorrected mistake each time they divide.
Why the 5′ to 3′ Direction Matters
DNA polymerase’s one-direction-only rule is not a limitation. It’s what makes proofreading possible. Because the enzyme adds nucleotides to the 3′ end of the growing strand, the most recently added base is always at the tip where the enzyme is working. If that base is wrong, the enzyme can immediately back up and remove it. If DNA were built in the opposite direction, removing a bad nucleotide from the growing end would destroy the chemical energy needed to add the next one, making efficient error correction impossible.
This directional constraint does create a complication: one strand of DNA (the lagging strand) runs in the “wrong” direction relative to the replication fork. DNA polymerase handles this by building the lagging strand in short segments called Okazaki fragments, each synthesized in the 5′ to 3′ direction. Another enzyme, DNA ligase, then seals the phosphodiester bonds between these fragments to create a continuous strand.