
Bacteria evolved restriction endonucleases to cut up invading phage DNA. We use them to cut vectors and inserts at sequences we choose.
They are the oldest tools in molecular cloning and they have not been replaced. Even in a lab that assembles everything with Gibson or Golden Gate, you still need an enzyme to linearize a vector, release an insert, or check a finished construct on a gel.
Most of the trouble they cause comes from three things: choosing an enzyme that cuts somewhere you did not expect, combining two enzymes whose buffers disagree, and pushing a reaction far enough outside its conditions that the enzyme starts cutting sequences it should ignore. This guide covers all three.
At a glance
- Two groups do the work: Type II enzymes cut inside their recognition site. Type IIS enzymes cut outside it, which is what makes Golden Gate possible.
- Double digests: Either digest sequentially with a cleanup in between, or use high-fidelity variants that share one buffer.
- Watch for: Glycerol above 5% in the reaction (which causes star activity), and methylation blocking a site that looks perfectly good on screen.
The types of restriction enzyme
Suppliers list hundreds of enzymes. They are grouped by their subunit structure, what cofactors they need, and where they cut relative to the sequence they recognize. Two groups do nearly all the work in cloning.
Type II enzymes
Type II enzymes are the standard enzymes for everyday cloning. EcoRI, BamHI and HindIII are typical examples. They usually recognize a palindromic sequence, four to eight base pairs long, and cut the backbone inside it.
The cut leaves either a short single-stranded overhang, often called a sticky end, or a blunt end. Because the enzyme cuts within its own recognition site, the site is reconstructed when two compatible ends are joined. This is why traditional restriction cloning leaves a scar at each junction: the sequence that let you make the join is still there afterwards.
Type IIS enzymes
Type IIS enzymes are a subgroup that cuts outside its recognition sequence. BsaI, BsmBI and BbsI are common examples.
They recognize an asymmetric sequence and cut at a fixed distance outside it. This means the four bases of the overhang are not part of the recognition sequence, so you can choose them. It also means that when the enzyme cuts, the recognition sequence leaves with the flanking DNA rather than staying in your part.
Two things follow: you can design overhangs that only pair with their intended partner, and because a correct join destroys the recognition site, the enzyme cannot cut the product again. Together these make ordered, multi-part assembly possible in a single tube, with no restriction site left behind at the junction (see Golden Gate Assembly).
Buffer compatibility and double digests
Restriction enzymes need particular salt concentrations, magnesium levels and pH to fold correctly and bind DNA. Suppliers ship each enzyme with a buffer that provides them.
The problem arrives when you want to cut a plasmid with two enzymes at once. If one needs a low-salt buffer and the other needs high salt, a single tube containing both will inhibit one enzyme or both.
There are two practical ways around this:
- Sequential digest: Cut with the first enzyme in its own buffer, clean up the reaction on a silica column to remove the salts, then cut with the second enzyme in its buffer. Reliable and cheap, but you typically lose 10% to 30% of your DNA in the cleanup.
- High-fidelity variants: Many suppliers sell engineered versions designed to work at full activity in a single universal buffer. This removes the intermediate cleanup and the DNA loss, at a higher price per reaction.
For a routine double digest where you have DNA to spare, sequential is fine. When your material is precious or your backbone is large and awkward to recover, the high-fidelity route usually costs less overall than losing a third of your vector.
Star activity
Star activity is what happens when an enzyme relaxes its specificity and starts cutting sequences that differ slightly from its recognition site. On a gel it shows up as extra fragments you cannot account for, or as a smear where you expected clean bands.
It is almost always caused by reaction conditions rather than by the enzyme:
- Too much glycerol: Enzymes are supplied in 50% glycerol so they do not freeze solid at -20 °C. Glycerol is also an organic solvent, and at high concentration it pulls away the water molecules that normally sit between the enzyme and the DNA and help keep recognition specific. The binding pocket becomes more tolerant, and the enzyme cuts sequences that come close to its site without matching it. Keep the enzyme solution below 10% of the total reaction volume, which keeps the final glycerol under 5%. In a 50 µL digest, this means no more than 5 µL of enzyme stock. Adding "a bit more enzyme" to a stubborn digest is the usual way people push a reaction past this limit.
- Other reaction conditions: Off-target cutting also increases with long incubations using standard enzymes, the wrong pH, or incorrect salt. If you need to extend an incubation, use a high-fidelity variant, which tolerates longer incubations better.
Plan it first
Most digest failures can be predicted from the sequence alone, before you set up the reaction.
If your gene of interest contains an internal EcoRI site and you did not notice, digesting your construct will cut the insert in two. Nothing about the setup will warn you; you find out on the gel, after the reagents are spent.
In Photo51, you can map every restriction site across your vector and insert at once, so an enzyme that cuts in three places instead of one is visible immediately (see also: Restriction Cloning in Photo51 and Virtual Gel Electrophoresis). You also get the expected fragment sizes for any combination of enzymes, giving you the exact reference to compare your gel against. A digest you cannot predict the pattern for is a digest you cannot interpret.
Troubleshooting
A PCR product with added sites will not cut
Symptom: You amplified a gene with restriction sites built into the 5' ends of the primers, and the digest barely works.
Cause: Restriction enzymes cut poorly (in some cases below 5% efficiency) when their recognition site sits right at the end of a linear fragment. The enzyme needs DNA on both sides to bind properly.
Fix: Add spacer bases. Most suppliers publish a cleavage-close-to-end table for their enzymes. Three to six extra bases on the 5' end of the primer, upstream of the restriction site, is usually enough. This has to be designed into the primer, so it is worth checking before you order rather than after.
Faint uncut bands alongside the expected fragments
Symptom: Your gel shows the fragments you wanted, plus fainter bands matching uncut or supercoiled plasmid.
Cause: The digest did not go to completion. Usually the enzyme has lost activity, or the buffer is wrong.
Fix: Check that you used the right buffer. Keep enzyme tubes on a pre-chilled cold block at the bench rather than in your hand, since these enzymes lose activity quickly at room temperature and after repeated freeze-thaw cycles. Then either use fresh enzyme or extend the incubation, and if you extend it, switch to a high-fidelity variant so you do not trade incomplete digestion for star activity.
A valid site will not cut, and the enzyme is fine
Symptom: The enzyme is active on a control, the buffer is right, and one particular site in your plasmid stays uncut.
Cause: Methylation. Standard E. coli cloning strains such as DH5-alpha protect their own DNA with methyltransferases. Dam methyltransferase adds a methyl group to the adenine in any GATC sequence, and if that sequence overlaps your restriction site, the methyl group sits in the enzyme's binding pocket and blocks it. An XbaI site (TCTAGA) followed by TC gives you TCTAGATC, which contains a GATC motif.
Fix: Check for overlaps of this kind at the design stage, since they are a property of the sequence and entirely predictable. If you find one, transform the plasmid into a dam-/dcm- strain such as JM110, prepare unmethylated DNA from it, and digest that.