Restriction Enzymes and How to Plan a Digest
6 min read · Updated June 8, 2026
Restriction enzymes are molecular scissors that cut DNA at specific sequences, and they remain a workhorse of cloning. This guide covers how they recognise their sites, the ends they leave, and how to plan a digest that gives you the fragments you want.
Recognition sites and cut positions
A restriction enzyme binds a short, usually palindromic recognition sequence — for example EcoRI recognises GAATTC — and cuts at a defined position within or near it. Some enzymes have degenerate recognition sequences described with IUPAC codes (for instance HinfI cuts at G^ANTC, where N is any base).
Sticky ends vs. blunt ends
- Sticky (cohesive) ends: a staggered cut leaves short single-stranded overhangs that base-pair with complementary overhangs, making directional ligation easier.
- Blunt ends: a straight cut leaves no overhang; blunt ligation is more flexible but less efficient and non-directional.
Planning a digest
- Scan your sequence for recognition sites so you know which enzymes cut, where, and how many times.
- Choose enzymes that cut your insert and vector compatibly — ideally leaving matching sticky ends.
- For a double digest, check the two enzymes share a compatible buffer and temperature.
- Predict the fragment sizes you expect to see on a gel before you run it.
Common pitfalls
Watch for enzymes that cut inside your insert (you'll lose or fragment it), star activity under suboptimal conditions, and methylation-sensitive sites that may not cut in DNA from certain hosts. Scanning the sequence first avoids most surprises.
Worked example: predicting fragment sizes for a single vs. double digest
Predicting fragment sizes before you run a gel is mostly bookkeeping: find each site, note exactly where the enzyme cuts within it, then measure the distances between cuts. Here is a full worked example on a 40-nt illustrative sequence (1-based positions, counting from the 5' end): AAAAAAAAAA GAATTC TTTTTTTTTT GGATCC CCCCCCCC — that is, positions 1-10, 11-16, 17-26, 27-32 and 33-40 respectively.
This sequence contains one EcoRI site (GAATTC, positions 11-16) and one BamHI site (GGATCC, positions 27-32). EcoRI cuts G^AATTC — between the 1st and 2nd base of the site — so here it cuts between positions 11 and 12. BamHI cuts G^GATCC the same way, between positions 27 and 28.
- EcoRI alone, linear DNA, one cut: two fragments of 11 nt (positions 1-11) and 29 nt (positions 12-40). Check: 11 + 29 = 40.
- BamHI alone, linear DNA, one cut: two fragments of 27 nt (positions 1-27) and 13 nt (positions 28-40). Check: 27 + 13 = 40.
- EcoRI + BamHI double digest, linear DNA, two cuts: three fragments of 11 nt, 16 nt (positions 12-27) and 13 nt (positions 28-40). Check: 11 + 16 + 13 = 40.
Overhang polarity, compatible ends, isoschizomers and Type IIS enzymes
"Sticky" isn't one thing. A staggered cut can leave a single-stranded overhang on the 5' end of each fragment (EcoRI's GAATTC leaves a 4-nt 5' overhang, AATT; BamHI's GGATCC leaves GATC; HindIII's AAGCTT leaves AGCT) or on the 3' end instead (PstI's CTGCAG leaves a 4-nt 3' overhang, TGCA; KpnI's GGTACC leaves GTAC). Blunt cutters like SmaI (CCCGGG) or EcoRV (GATATC) leave no overhang at all. Two ends will only ligate if the overhang sequence and its polarity match exactly — "both sticky" is not enough.
Because compatibility depends on the overhang, not the enzyme's name, different enzymes with different recognition sequences can still produce identical, ligatable ends. BamHI (G^GATCC), BglII (A^GATCT), BclI (T^GATCA) and the four-cutter Sau3AI (^GATC) all leave the same 5'-GATC overhang, so an insert cut with one of them can be ligated into a vector cut with another. The trade-off is that the resulting junction sequence is a hybrid of the two recognition sequences and, for the two 6-base cutters, usually matches neither original site — so plan on that junction no longer being cuttable by either enzyme.
Two enzymes that recognise the exact same sequence are isoschizomers only if they also cut at the same position; if they recognise the same sequence but cut differently, they're neoschizomers (SmaI/XmaI, above, is the standard example). A separate category worth knowing about is Type IIS enzymes such as BsaI or BsmBI, used in Golden Gate-style assembly: they cut a fixed distance outside their own recognition sequence, so the recognition site tells you where the enzyme binds but not where it cuts or what overhang it leaves — that has to be looked up per enzyme rather than read off the site itself.
Mistakes that show up in double digests and ligations
- Forgetting the circular-vs-linear fragment-count rule: N cut sites in a circular plasmid give N fragments, but N cut sites in a linear piece of DNA give N+1 fragments (see the worked example above for the linear case). Applying the linear rule to a plasmid, or vice versa, is a frequent source of an unexpected band count.
- Assuming a shared overhang means the recognition site survives ligation: as covered above, ligating a BamHI end to a BglII end creates a hybrid junction that generally isn't a site for either enzyme any more — don't plan a later step around re-cutting it without actually checking the resulting sequence.
- Scanning only the insert or only the vector for cut sites, then assuming that result holds for the final construct: the ligation junction can accidentally create a brand-new recognition site, or, as above, destroy an expected one. Re-scan the assembled sequence.
- Reading a Type IIS enzyme's recognition sequence and expecting the cut and overhang to sit inside it the way EcoRI's or BamHI's do — for enzymes like BsaI or BsmBI the cut is offset outside the recognition sequence, so the site alone doesn't tell you the overhang.
Frequently asked questions
What is the difference between sticky and blunt ends?
Sticky ends have short single-stranded overhangs from a staggered cut, which base-pair to aid ligation. Blunt ends come from a straight cut with no overhang and ligate non-directionally.
How do I know if an enzyme cuts my sequence?
Scan the sequence for the enzyme's recognition site. A restriction site finder reports every match, its position and the cut site, including degenerate recognition sequences.
If a plasmid has two sites for the same enzyme, how many fragments does the digest give?
It depends on whether the DNA is circular or linear. Two cuts in a circular plasmid release two fragments, because the two ends created by each cut are already joined around the circle. Two cuts in a linear piece of DNA give three fragments, because the original two ends of the linear molecule are also fragment ends. Forgetting this is a common reason a digest shows one band more or fewer than expected.
Are all sticky ends compatible with each other?
No. Only ends with the identical overhang sequence and the same polarity (5' or 3') will base-pair and ligate. Different enzymes can still produce compatible ends — BamHI, BglII, BclI and Sau3AI all leave a 5'-GATC overhang despite recognising different sequences, so a BglII-cut insert can be ligated into a BamHI-cut vector. Just be aware that the resulting junction usually matches neither enzyme's original recognition sequence, so you generally can't cut it again with either enzyme afterwards.
What's the difference between isoschizomers and neoschizomers?
Isoschizomers recognise the same sequence and cut at the same position, so they produce identical fragments. Neoschizomers recognise the same sequence but cut at a different position within it. SmaI and XmaI are the classic example: both recognise CCCGGG, but SmaI cuts it blunt (CCC^GGG) while XmaI cuts off-centre (C^CCGGG), leaving a 4-base 5' overhang.
Related references
Common restriction enzymes: recognition sites, cut positions, NEB buffer activity, star activity and an interactive double-digest buffer finder.
Selection markers, mechanisms and working concentrations for cloning.
Reference table of common cloning and protein expression vectors with backbone size, origin of replication, copy number, selection marker, promoter and fusion tags.