How to Find an Enzyme That Cuts Your Plasmid Exactly Once

10 min read · Updated August 11, 2026

You need to open a plasmid at one place: to linearise a backbone before an assembly, to make a template for in vitro transcription, to drop in a cassette, or just to check a prep is what you think it is. The requirement sounds simple — an enzyme that cuts exactly once — and the list of those is usually shorter and less useful than expected.

This guide covers how to read a single-cutter list, the scanning mistake that silently removes real sites from it, and the questions that decide which of the candidates is actually the right one. The worked example is pUC19, so you can reproduce every number.

Unique is necessary, not sufficient

Scan pUC19 (2,686 bp) against a curated set of 49 common enzymes and you get 109 sites in total. Fourteen of those enzymes cut exactly once: EcoRI, BamHI, HindIII, SalI, PstI, SmaI, KpnI, SacI, XbaI, NdeI, ScaI, SphI, SbfI and ApoI.

Fourteen options sounds comfortable until you look at where they cut. Twelve of the fourteen sites fall between positions 396 and 447 — a 52 bp window, because that is the polylinker, and a polylinker is precisely a dense cluster of sites that occur nowhere else. For the purpose of cloning something into the MCS that is exactly what you want. For the purpose of opening the plasmid somewhere else, those twelve are one option, not twelve.

Outside that window there are two: NdeI at position 183 and ScaI at position 2,177. If your requirement is "linearise this backbone away from the insert site", the real answer for pUC19 is a choice between two enzymes, and it is worth knowing that before you plan around a list of fourteen.

A linear scan of a circular plasmid quietly loses sites

This one costs people real time, and it produces a wrong answer that looks like a right one.

A plasmid file is a linear string with an arbitrary start position, chosen by whoever deposited it or by whichever tool last wrote it out. The molecule has no start. A recognition site that happens to straddle that arbitrary join exists on the plasmid and does not exist in the string, so a scanner that treats the sequence as linear will not find it.

Here is the same pUC19 rotated so that its EcoRI site sits across the join — a rotation, nothing added or removed, the same molecule. Scanned as circular: 109 sites, 14 single cutters, EcoRI and ApoI both present. Scanned as linear: 107 sites, 12 single cutters, and EcoRI and ApoI have moved into the list of enzymes that do not cut pUC19 at all.

"EcoRI does not cut pUC19" is a memorably wrong sentence, and the only thing that produced it was a checkbox. The error is worse in the other direction too: an enzyme that cuts once inside the string and once across the join is a genuine double cutter reported as a unique one, which is a digest that gives you two bands you did not plan for.

So set circularity for anything circular, always. It costs nothing, and the failure mode is silent in both directions.

Where the cut lands is the real question

Once you have candidates that genuinely cut once, the choice between them is about position, and it is worth annotating the plasmid rather than reasoning from coordinates.

On pUC19, ScaI cuts at 2,177, which is inside the AmpR (bla) coding sequence — that gene runs 1,626-2,486 on the minus strand. Linearising there is completely fine if the product is a template you will transcribe or a fragment you will assemble into something else. It is not fine if you plan to religate and transform, because you have cut the resistance marker in half.

The same logic applies to every functional element. A cut inside the origin will not replicate; a cut inside your promoter or your insert defeats the purpose; a cut inside a region you are about to use as a homology arm removes the homology. Annotate first, then choose.

Fragment sizes matter for a different reason. If the point of the digest is diagnostic — confirming a prep is the plasmid you think it is — a single cut gives one band and tells you the total length and not much else. Two well-separated cuts give two bands whose sizes are a much stronger claim about identity and orientation, which is why a diagnostic digest and a linearisation digest are different design problems.

The enzymes that do not cut are half the answer

The list nobody looks at is often the one that solves the problem. Twenty-four of the 49 enzymes scanned do not cut pUC19 anywhere: NotI, XhoI, SpeI, NcoI, NheI, EcoRV, HpaI, StuI, ApaI, BglII, ClaI, MluI, NsiI, AflII, AgeI, AscI, AvrII, BspEI, BstBI, FseI, PacI, PmeI, SwaI and EagI.

That list is what you consult when you are adding a site rather than using one — putting a cloning site on the end of a PCR product, choosing an enzyme for a downstream step that must not touch the backbone, or picking the pair for a directional cloning strategy. An enzyme that cuts your insert once and your vector never is exactly what a clean two-enzyme cloning plan needs.

It is also the fastest way to notice that your favourite enzyme is unavailable before you have designed around it.

Two things that make a site list less literal than it looks

Degenerate recognition sequences. Not every enzyme reads a fixed six bases. ApoI recognises R^AATTY — a purine, AATT, then a pyrimidine — so it matches GAATTC among other things, which is why its single site on pUC19 sits at position 396, the same position as EcoRI's. Two enzymes, one location. When a site list shows two enzymes at an identical coordinate, that is usually what is happening rather than a duplicate row.

A curated list is not every enzyme. Scanning 49 common enzymes is the right default for a design decision, because those are the ones in the freezer. It is not a statement about the thousands of characterised restriction enzymes, so "not in the non-cutter list" and "cuts your plasmid" are different claims. If a specific unusual enzyme matters, check it by name rather than inferring from its absence.

Two more that are not visible in any site list at all: methylation and star activity. Dam and Dcm methylation of plasmid DNA grown in standard E. coli blocks some enzymes entirely, and long digests with excess enzyme can produce cuts at near-cognate sites. A site list tells you where the sequence matches; it does not tell you what the enzyme will do to DNA out of your particular strain.

A working procedure

Annotate the plasmid first, so you know which regions must survive the cut.
Scan with circularity switched on. Every number changes if you get this wrong, in both directions.
Read the single-cutter list together with the positions, not on its own — a dozen unique cutters in one polylinker is one usable location.
Cross out the candidates whose position lands inside something you need: the resistance marker, the origin, your insert, a homology arm.
Check the survivors against what is actually in your freezer and against buffer compatibility if the digest is a double.
Check for methylation sensitivity before committing to an enzyme, especially for plasmid DNA from a standard cloning strain.
If nothing suitable cuts once, invert the problem: use the non-cutter list to pick a site you can add by PCR, or use a nicking or Type IIS strategy instead.

The tool

SeqBench's Restriction Sites tool scans a sequence against its curated common-enzyme set — or against a named subset you supply — and returns each enzyme's recognition sequence, cut offset, number of cuts and every position, alongside three summary lists that are the useful part here: the enzymes that cut exactly once, the ones that cut at all, and the ones that do not cut. The circular option is what makes it correct for a plasmid, and it is off by default because not everything you paste is one.

Every count in this guide came from running it on pUC19 (GenBank L09137), including the rotated-plasmid comparison, so all of it is reproducible in a browser tab. For choosing a pair of enzymes and a buffer, the double-digest tool is the better fit, and for what the resulting fragments would look like on a gel, the virtual gel takes it from there.

Frequently asked questions

How do I find an enzyme that cuts my plasmid only once?

Scan the whole plasmid with circularity switched on and read the single-cutter list together with the cut positions. On pUC19, 14 of 49 common enzymes cut exactly once — but 12 of those 14 sites sit in a 52 bp window in the polylinker, so as locations rather than enzymes there are only three choices. Then eliminate any candidate whose position falls inside something you need intact.

Why does my software say an enzyme does not cut when I know it does?

Almost always because the plasmid was scanned as a linear sequence. A plasmid file's start position is arbitrary, and a recognition site straddling that join exists on the molecule but not in the string. Rotating pUC19 so its EcoRI site crosses the join and scanning it linearly moves both EcoRI and ApoI out of the single-cutter list and into the list of enzymes that do not cut it at all — 107 sites instead of 109. Switch circularity on for anything circular.

Can a linear scan also report a double cutter as unique?

Yes, and this is the more dangerous direction. An enzyme with one site inside the string and one across the arbitrary join is a genuine two-cutter that a linear scan reports as cutting once — so a digest you planned as a linearisation comes back as two fragments. The fix is the same: scan circular sequences as circular.

How do I pick between several unique cutters?

By position, not by preference. Annotate the plasmid first and eliminate any cut that lands inside the resistance marker, the origin, your insert, or a region you are about to use as a homology arm. On pUC19, ScaI cuts at 2,177, which is inside the AmpR coding sequence at 1,626-2,486 — fine for making a template, wrong if you intend to religate and select on ampicillin. After position, sort by what is in your freezer and by buffer compatibility.

Why do two enzymes show a site at exactly the same position?

Usually because one of them has a degenerate recognition sequence that includes the other's. ApoI reads R^AATTY, which matches GAATTC among other sequences, so its single site on pUC19 coincides with EcoRI's at position 396. It is not a duplicated row; it is two enzymes that genuinely cut in the same place, though not necessarily leaving the same ends.

Which enzymes do not cut my plasmid, and why would I want to know?

That list is what you use when adding a site rather than using one. Twenty-four of the 49 common enzymes scanned do not cut pUC19 at all, which makes any of them a safe choice to append to a PCR product for directional cloning, or to use in a downstream step that must leave the backbone untouched. Checking it early also saves designing a strategy around an enzyme that turns out to cut your vector three times.

Does a single cutter give me a useful diagnostic digest?

Not really — one cut gives one band, which confirms the total length and little else. A diagnostic digest wants two or more well-separated cuts, because the pattern of fragment sizes is a much stronger claim about identity and orientation than a single number. Treat linearisation and diagnostic digestion as separate design problems that usually want different enzymes.

Can I trust a site list completely?

It tells you where the sequence matches, which is not the same as what the enzyme will do. Two things sit outside any site list: Dam and Dcm methylation from a standard cloning strain, which blocks some enzymes entirely at otherwise perfect sites, and star activity in long digests with excess enzyme, which produces cuts at near-cognate sequences. Also remember a curated set covers the enzymes people actually own, so an enzyme's absence from the non-cutter list is not evidence about it.